Skip to content

AI-powered virtual executive team — a single coherent executive persona backed by 8 specialist agents (FastAPI + Next.js).

feat(integrations): say it is an AI on mail and chat the Executive sends (#449)

Rufusb5677f48d97a
diff --git a/.env.example b/.env.example
index 9091ffbb..0cdd40e3 100644
--- a/.env.example
+++ b/.env.example
@@ -244,6 +244,17 @@ SLACK_APP_TOKEN=xapp-your-app-token
# signed in below (see EMAIL_PROVIDER). No default: the Executive must never
# operate from a fallback address because EXEC_EMAIL_ADDRESS silently fell through.
+# AI disclosure: mail and chat the Executive sends end with a line saying it
+# is an AI assistant (orchestrator/disclosure.py). On by default. The texts
+# replace the built-in wording and must say "AI", or the built-in one is used.
+# AI_DISCLOSURE=true
+# AI_DISCLOSURE_MAIL_TEXT=
+# AI_DISCLOSURE_CHAT_TEXT=
+# Where each person's own switches start for replies sent as them under Act as
+# me: those Handle it for me sends on its own (default on), and drafts they
+# approve and send (default off). Each person can change theirs.
+# AI_DISCLOSURE_HANDLE_IT=true
+# AI_DISCLOSURE_APPROVED_SENDS=false
# How often the email poller checks for new messages (seconds, default 60)
EMAIL_POLL_INTERVAL_SECONDS=60
# Channel access (Email/Telegram/Discord) is roster-driven: only senders
diff --git a/packages/core/openexecutive/api/routes/delegation.py b/packages/core/openexecutive/api/routes/delegation.py
index 9b7b9823..0eb35145 100644
--- a/packages/core/openexecutive/api/routes/delegation.py
+++ b/packages/core/openexecutive/api/routes/delegation.py
@@ -257,6 +257,24 @@ class HandleItUpdate(BaseModel):
mode: str | None = None
+class DisclosureOut(BaseModel):
+ """The AI line on replies sent as the caller (delegation.disclosure):
+ on those Handle it for me sends on its own, and on drafts they send.
+ ``enabled`` is the install's AI_DISCLOSURE: off, neither is added."""
+
+ enabled: bool
+ handle_it: bool
+ approved_sends: bool
+ line: str
+
+
+class DisclosureUpdate(BaseModel):
+ model_config = ConfigDict(extra="forbid")
+
+ handle_it: bool | None = None
+ approved_sends: bool | None = None
+
+
class HandledReplyOut(BaseModel):
decision_id: int
sent_at: str
@@ -282,6 +300,8 @@ class DelegationOut(BaseModel):
handle_it: HandleItOut | None = None
# Absent on a backend that predates per-setting training.
training: TrainingOut | None = None
+ # Absent on a backend that predates the AI line on replies.
+ disclosure: DisclosureOut | None = None
# The principal's "Let team members use Act as me", while the install
# allows it; None for anyone else.
team: TeamOut | None = None
@@ -544,10 +564,22 @@ async def _state(person: Person) -> DelegationOut:
inbox=_inbox_out(_person_id(person)),
handle_it=_handle_it_out(_person_id(person)),
training=_training_out(_person_id(person)),
+ disclosure=_disclosure_out(_person_id(person)),
team=_team_out(person),
)
+def _disclosure_out(person_id: int) -> DisclosureOut:
+ from openexecutive.delegation import disclosure
+ from openexecutive.orchestrator.disclosure import enabled
+
+ switches = disclosure.get(person_id)
+ return DisclosureOut(
+ enabled=enabled(), handle_it=switches.handle_it, approved_sends=switches.approved_sends,
+ line=disclosure.LINE,
+ )
+
+
def _handle_it_out(person_id: int) -> HandleItOut:
from datetime import UTC, datetime
@@ -901,6 +933,27 @@ async def update_delegation_handle_it(request: Request, body: HandleItUpdate) ->
return await _state(person)
[email protected]("/delegation/disclosure", response_model=DelegationOut)
+async def update_delegation_disclosure(request: Request, body: DisclosureUpdate) -> DelegationOut:
+ """The caller's own switches for the AI line on replies sent as them.
+ Turning one on only adds the line, so it needs no signed sign-in;
+ turning one off does, so nobody else can take it off their mail."""
+ from openexecutive.delegation import disclosure
+
+ changes = {k: v for k, v in body.model_dump().items() if v is not None}
+ person = _signed_caller(request, narrowing=all(changes.values()))
+ person_id = _person_id(person)
+ before = disclosure.get(person_id)
+ changed = {k: v for k, v in changes.items() if getattr(before, k) != v}
+ if changed:
+ disclosure.set_(person_id, **changed, updated_by=f"person:{person_id}")
+ _audit(
+ "delegation_disclosure_changed", f"AI line on replies changed for person {person_id}",
+ {"person_id": person_id, **changed},
+ )
+ return await _state(person)
+
+
@router.put("/delegation/training", response_model=DelegationOut)
async def update_delegation_training(request: Request, body: TrainingUpdate) -> DelegationOut:
"""Put Act as me's settings in or out of training. Putting Replies or
diff --git a/packages/core/openexecutive/architecture/architecture-facts.yaml b/packages/core/openexecutive/architecture/architecture-facts.yaml
index 8d523c51..e6343bff 100644
--- a/packages/core/openexecutive/architecture/architecture-facts.yaml
+++ b/packages/core/openexecutive/architecture/architecture-facts.yaml
@@ -1331,6 +1331,33 @@ routing:
the names from the handler map ("unknown tool — skipped"). The alert
triage backstop matches: alerts.preferences.resolve_channels strips
department_channel and company_broadcast in solo.
+ ai_disclosure_invariant: |
+ What the Executive sends to people says it is an AI assistant, added by
+ plain code, never the prompt (orchestrator.disclosure). Mail from its
+ own mailbox: the MCP gateway appends the mail line to every Gmail send
+ or draft and every Outlook send, reply, forward or draft
+ (mcp_gateway._with_disclosure) after every outbound gate passed and only
+ on the arguments sent to the server, so gates, outbound-context records
+ and audit rows see the model's words. Skipped when every recipient is
+ the principal, when the body already has it, and for send_draft_message
+ (the draft got it when made). Only a last shown line counts as already
+ there. HTML that could hide what follows (disclosure.hides_what_follows)
+ is refused; the roster acknowledgement to a new sender gets
+ BARE_MAIL_LINE, which names nobody. Chat: the shorter line on the Executive's
+ Slack / Telegram / Discord / Google Chat replies and on send_slack_dm,
+ send_discord_dm, send_telegram_message and broadcast_tools._dispatch,
+ once per conversation per 24 hours (in memory; a restart shows it
+ again), applied at the send call only so the anti-spam guard and session
+ history see the text without it. AI_DISCLOSURE (default on) turns it
+ off; AI_DISCLOSURE_MAIL_TEXT / AI_DISCLOSURE_CHAT_TEXT reword it and are
+ ignored unless they say "AI". Fails open. Sends as a person (Act as me)
+ carry delegation.disclosure.LINE by that person's own switches
+ (delegation_disclosure: handle_it default AI_DISCLOSURE_HANDLE_IT=true,
+ approved_sends default AI_DISCLOSURE_APPROVED_SENDS=false), written into
+ the reply when it is drafted (inbox._consider, follow_ups) so card,
+ draft and sent mail match. PUT /delegation/disclosure changes them;
+ turning one off needs a signed caller. tests/conftest.py sets
+ AI_DISCLOSURE=false for the suite; test_ai_disclosure.py turns it on.
outbound_antispam_invariant: |
Every Per-Person DM (`send_slack_dm` / `send_discord_dm` /
`send_telegram_message`) passes through a deterministic anti-spam guard
diff --git a/packages/core/openexecutive/architecture/prebuilt/agents.json b/packages/core/openexecutive/architecture/prebuilt/agents.json
index e7444024..da51a89f 100644
--- a/packages/core/openexecutive/architecture/prebuilt/agents.json
+++ b/packages/core/openexecutive/architecture/prebuilt/agents.json
@@ -1,7 +1,7 @@
{
"section_id": "agents",
"title": "Agent Council",
- "markdown": "The Executive (`openexecutive.orchestrator.executive.Executive`) is a single coherent persona that fans out to specialist sub-agents via Anthropic tool use. Each specialist subclasses `openexecutive.agents.base.BaseAgent`, carrying a `name`, `domain`, and `model`, and exposes `analyze(...)`, which wraps the question in user-turn context and calls Claude through the per-call provider. That context is per call and never cached: `<company_stage>` (the stage from the company profile — specialists never see the profile itself), in solo mode `<principal_role>` (what kind of principal the advice is for, in plain words, with their title and remit — `router.principal_role_context`), `<department_memory>`, `<past_decisions>`, `<failure_cases>`, `<relevant_knowledge>` and `<conversation_context>`, outermost first. The specialist's own system prompt is a static domain constant under one `cache_control` block, so it stays byte-identical across companies and profile edits. A chat turn takes the stage from its session's profile (the one the Executive reasons over, and the one an eval scenario injects); workflow steps and the MCP server's `consult_specialist` read it fresh from the profile on disk. The role tag follows the turn's mode: a chat turn pins the role on its session with the mode (`pin_turn_principal_role`: the session's `principal_role` override, else the workspace settings) and sends it in solo and nothing in team, so the org block, every specialist in the turn and any workflow the turn starts see the same role; callers with no chat turn read it fresh (`load_principal_role`: the current session's mode and role, else the workspace's), and a failed read sends no tag rather than failing the consult. The CFO, CSO and sales prompts key their advice on it: for a bootstrapped, self-funded or solo company they drop venture fundraising benchmarks and work from cash, and raise only if the founder wants to.\n\nRouting uses one `consult_specialist` tool. Beside it the Executive carries two skill tools, `search_skills` and `load_skill` (skills are shown to users as **playbooks**, and every playbook belongs to a workflow: one that follows it, or its own Quick workflow). Chat finds and reads playbooks but never creates or changes one: work worth repeating is offered as a workflow through `draft_workflow`, with the method written into its steps, because workflows read playbooks at run time (approved scheduled ones included) and this loop also runs on inbound email and chat, so only a person can change what a workflow follows. A hidden built-in is invisible to search and load; each `search_skills` hit lists the `workflows` that follow it, Quick ones included, and the persona routes \"the finished deliverable\" to that workflow and quick or partial asks to the playbook inline; Tool calls made while proactive work is tagged (scheduler dispatch, reflection, research, alert review) feed the Attunement outcome ledger. Goals have three tools: `list_department_goals`, `update_department_goal` (progress the user reports) and `create_goal` (a new goal the principal states, filed under an area by slug or title and creating the area when none matches — see Org). `create_goal` runs only for the principal on a verified surface, and no unattended run is offered it: the tools in `schedule_tools.UNATTENDED_WITHHELD_TOOLS` are dropped from reflection's and research's toolkits by `unattended_toolkit`, and from the chat loop when its Session is marked `unattended` — the scheduler's PROACTIVE TRIGGER run, whose prompt quotes stored intent text; a call the model emits anyway gets an error tool_result. `record_decision_outcome` records how a past decision turned out — the weekly review lists decisions older than 30 days with no outcome as `[decision N]` — writing `decisions.outcome` (`episodic.update_decision`) and auditing the rationale; it has `create_goal`'s gate (the principal on a verified surface only; everyone else gets an audited refusal) and is in `UNATTENDED_WITHHELD_TOOLS` too. A missing or unknown id returns the decisions still waiting on an outcome, so the model can retry with the right one. Its schema is static; the solo persona has one line on it, and team mode learns it from the tool description, so the team persona is byte-identical. Attunement adds `list_open_loops` (what a person owes — everyone's for the principal only in a private conversation, otherwise the caller's own) and `close_open_loop`, which only closes a loop when the turn's resolved speaker is the principal or the loop's owner — an unrostered sender, a signed-in web user not on the roster, and unattended sessions are refused. `assign_open_loop` opens one on request (\"ask Ben to send me the Q3 numbers by Friday\") for the principal or a rostered teammate on a verified surface — never an email, a turn private to the principal or an unattended run, which are also withheld the tool — and only when `task` appears verbatim in what the speaker typed this turn, at most 5 per turn; that turn's extraction pass then opens no second loop for the same owner. The roster write tools (`upsert_person`, `archive_person`, `set_department_head`) are the principal's alone, and only on a surface that verified the speaker — the web chat, the principal's own Slack or Discord, or a private Telegram chat when `TELEGRAM_WEBHOOK_SECRET` is set to a value Telegram can send (Telegram identifies the chat rather than the sender, and without that secret its webhook accepts anyone's update). A teammate, an email (even from the principal's address), Google Chat, the CLI, the MCP server and unattended runs get `status: refused` (audited, and no action chip), because a team member's People row is also the web sign-in and outbound-email allow-list. `upsert_person` takes `kind` (`team` / `contact`; a new person defaults to `contact` in solo mode, an update keeps the current kind) and `email_aliases` (other addresses the person writes from — the full list, replacing the current one; they match mail and may be emailed but never sign in, and one already on someone else is refused), and after every write any pending roster request whose sender now matches someone is closed and its held messages replayed. `resolve_roster_request` answers a **roster request** — someone not on the People list who wrote in and is waiting (see Org): `approve` (a new person; `kind` defaults to team for a company-domain sender, else contact, when the principal did not say), `link` (someone already listed, by `person_id`; `replace_channel_id` only when the principal said so) or `decline`. It has the roster tools' gate (the principal on a verified surface; audited refusal otherwise), is in `UNATTENDED_WITHHELD_TOOLS`, and answers only an id in `Session.trusted_roster_request_ids` — the ids the server printed that turn in a `<roster_requests>` block (`briefing.context.format_roster_requests_for_prompt`, up to 10 pending requests), which `render_and_trust` appends to the `<briefing>` digest on the principal's own verified turn only and records on the session in the same step, so an id the model invents or reads out of a message adds nobody. Each line is server-derived — id, channel, address or account id, the sender's name sanitised and marked unverified, a company-domain hint, the message count — and never what they wrote. The block rides in the **user turn**, like the digest, so the cached system blocks are unchanged; the tool definition is static, and the new tool only extends the sorted tool list. The web chat route resets the trusted set every turn, and a resolved id is removed from it. and `list_people` returns both kinds with `kind` — contacts only on the principal's own verified turn (they are private to the principal; anyone else gets the team as if no contact existed) — while `lookup_person` searches the team only. `ask_about_person` reads a person's peer memory fully only on their own verified turn; anyone else's answers only to a verified principal or teammate, and only when that person turned on \"Share my work style with the team\", framed to how they work and never their view of a colleague, and otherwise returns the same \"hasn't chosen to share\" result whether or not anything is known. A teammate's own mail with the Executive is theirs the same way: every mailbox call the model makes through the MCP gateway (Gmail or Outlook, by `call_tool`, from a script, or a workflow step run from chat) passes through `orchestrator.teammate_mail`. A search excludes every other teammate's address server-side and is filtered again entry by entry; any other call that names a message or thread (read, attachment, reply, forward, label) must use an id a filtered search returned to the same reader in this conversation (`Session.mail_ids_seen`) or the inbound email the turn answers; reads are filtered message by message, so a message naming another teammate is left out unless the reader is in its From / To / Cc; a reply that would quote such a thread, or an Outlook forward, reply or attachment read of such a message, is refused; and a conversation in which a teammate other than the principal was shown their own mail becomes theirs alone, as Act as me's do. Outside mail and the reader's own are untouched; an email turn reads as its sender only when Gmail authenticated it, and any other unverified reader sees no teammate's mail, a background run the model drives reads as the principal and a workflow step outside any session as nobody, an Outlook inbox rule that forwards or names another teammate is refused, and the poller and the brief, which read the mailbox themselves, are unaffected. Their own chats are theirs too: every audit row of a teammate's verified turn is theirs (`people_tools.teammate_turn_owner`, see Audit; the principal sees the actions, never the words), and a teammate's web chat or DM opens for them alone (`chat._session_access`), the principal included. The same verified-principal rule, as `people_tools.contacts_reachable_now()`, decides whether a turn may reach a **contact** at all: `message_person`, the raw DM tools, `create_calendar_event` / `create_instant_meeting` attendees and the MCP gateway's Gmail / Calendar / Drive / OneDrive gates treat a contact exactly like an unknown person on every other turn, so no refusal reveals that one exists. **The untrusted-content policy** (`orchestrator.content_trust`) is one rule for text the principal did not write — inbound email, chat, attached files, watched pages — answering three questions for every surface. *Who is speaking*: the principal only when the surface proved it (`principal_speaking`: the web chat as the principal — never a signed-in email on nobody's People entry —, the CLI, their own Slack/Discord or a secret-verified Telegram chat, or their own address with Gmail's `dmarc=pass`); never a teammate, a stranger, Google Chat, the MCP server or an unattended run. *What may become memory*: episodic extraction runs only then, and only on the words outside `<untrusted_content>` blocks; peer memory and open loops are recorded under the speaker's own name, never as the principal's. *What may trigger tools*: every write on the principal's authority already refuses others in its handler; the tools with no such check that change the install for every later turn — `content_trust.PRINCIPAL_ONLY_TOOLS`: `load_mcp_server`, `list_saved_tools`, `run_python_job` and `run_script` — are not offered unless the principal is speaking on an interactive surface (not email, even authenticated), and a call the model emits anyway gets an error tool_result and a `refused: not_principal` audit row. So an inbound email, a teammate's turn, Google Chat and the scheduler's proactive trigger each get a stable tool list of their own without it. *How it is labelled*: outside text goes into the user turn inside `<untrusted_content source=… author=… author_verified=…>` with a fixed notice that it carries no authority; compatibility forms are folded (NFKC), control characters go, and the tag's name is renamed wherever the text carries it — any case, spacing or entity-escaped form — so the text cannot end its block and speak outside it.\n\nA turn about the principal's private mail (`Session.private_to_principal`) reaches the principal only. It is neither offered nor allowed to run `schedule_tools.PRIVATE_TURN_WITHHELD_TOOLS` — everything that reaches someone else, publishes where the team reads, or starts work outside the turn: the broadcasts, meeting bookings and cancellations, `run_workflow` and research, follow-ups and workflow nudges, watches, artifacts, goal updates, saved workflows, skill drafts, and `load_mcp_server` (it connects to any URL, which can carry the turn's content). Through `search_tools` and `call_tool` it reaches only `schedule_tools.PRIVATE_TURN_MCP_TOOLS`: Google Workspace reads (mail, calendar, and Drive, Docs and Sheets search and file reads, with their Microsoft 365 twins; OneDrive's `download-bytes` only for a file's content) and the Gmail send and draft the gateway narrows to the principal. A search shows it nothing else, and a call to anything else (Google Chat, a Docs or Drive write, `manage_event`, another server) is refused (`schedule_tools.private_turn_withholds`). A call the model emits anyway, MCP tools included, gets an error tool_result and writes an audit row, and every audit row the turn writes is the principal's alone. On every turn the Executive knows what is connected without searching: block 0's *Connected Systems* section names the connected Google services, other MCP servers, the messaging channels, Notion sync, Confluence sync (when on) and Act as me, and pins the Google tool names it calls directly with `call_tool`. Both the service list and the pinned names are narrowed to what workspace-mcp actually loaded (`WORKSPACE_MCP_TOOLS`, `WORKSPACE_MCP_TOOL_TIER`), so a service left out is not named and its tools are not pinned — being told a tool is already available when it is not in the index costs a failed call the model cannot explain to itself. `search_tools` is for anything not listed. When work repeats over many items, the Executive can also call `run_script` (`workflows/step_script.py`, offered when the gateway runs, unless `CHAT_SCRIPTS=false`, and only on the principal's own turns — `PRINCIPAL_ONLY_TOOLS`, withheld on an unattended pass too, because it spends the turn's whole call budget on code the model just wrote): one short Python script, run in a sandboxed Monty worker with no files, network or environment, whose only reach is the gateway tools `call_tool` could reach (those `search_tools` has returned; the gateway keeps that set for its own process-wide session, not per conversation). Each call it makes is a `call_tool` in every way that matters — the gateway's discovery, deny-list and recipient gates, Take the lead's gate on an unattended run, and the same action chip and `tool_invocation` row (marked `via: run_script`, `ok: false` when it raised), as each call happens; the script itself gets a `kind: script` row with its source and result. Its wall clock is half the turn's deadline, so a long script ends with the list of calls it made rather than being cut off with the turn, and a stopped turn closes it and its worker. A turn private to the principal is not offered it, and the Act as me lockdown withholds it whole. Besides gateway tools, a chat script may call a fixed list of the Executive's own tools (`step_script.CHAT_OWN_TOOLS`: reads, alerts, open loops, goals, the watch list, follow-ups, `message_person`, `upsert_person` — nothing that broadcasts, books calendars, edits memory or skills, or starts workflows or research), each only when the turn is offered it and through the turn's own handler (Take the lead's gate included), with the same chip and a `kind: skill`, `via: run_script` audit row. A direct call needs a `tool_use` each, so a turn's scripts share a budget instead: `CHAT_SCRIPT_MAX_CALLS` (default 200) calls in all, and per-turn caps on own tools that could otherwise act in bulk (`CHAT_OWN_TOOL_CAPS`: `message_person` 10 — messaging the whole roster would be a broadcast — `upsert_person` 25, `create_alert` 50); a refused call tells the model to say what is left. A call stopped mid-flight by the script's clock or the turn still gets a `cancelled` audit row, since it may have run. Because every round re-reads the conversation, the loop also nudges at the moment of choice: after a round whose results include a list of five or more items (`step_script.lists_many`), with `run_script` on offer and not used that round, it adds the constant `FANOUT_HINT` as a text block in that round's user message (never a cached block), once a turn, pointing the per-item work at one built tool. For library work on files (PDFs, spreadsheets, charts, images) the chat also has `run_python_job` (see Workflows → Python jobs): Python with a fixed library set in a Deno-isolated WebAssembly sandbox, fed Knowledge files by name, the message's own attachments as sent, or text it already holds, and able to keep a job that worked as a Custom tool, with result files linked for the principal to download; it is offered only on the principal's own turns and only where the sandbox is installed (not offered otherwise, so the tool prefix stays fixed per process). Agent Activity shows each run as one \"Built tool\" card (debug event `script_run`: ok, the calls it made, how many, how long, and the kept or reused tool's name). The persona's tool-building section is present only when chat scripts can actually run (the setting and Monty installed). `TOOL_BUILDING_ADDENDUM` in the cached persona tells the Executive when to build a tool — a job that looks something up and then checks or acts on each result, paged results, more than about ten items, or an answer that needs three or more reads (find, open, read; the tool returns only what answers the question, not whole documents, and only what is about to be edited or quoted at length is read in full), while a handful of items already named go as direct calls in one turn — and never to say it has none, and to talk about it as building a tool or doing it in one go. On the principal's own verified, interactive turns, a script that works can be saved as a named tool (`save_as`; elsewhere the script runs but is not saved) and run again with `run_script(tool=…, inputs=…)`; `list_saved_tools` (a constant definition, in `PRINCIPAL_ONLY_TOOLS` with `load_mcp_server`, so offered only while the principal is speaking on a verified, interactive surface; `run_script(tool=…)` is refused on any other turn) lists them, and each call a saved tool makes is a gateway call like any other (see Workflows → Saved tools). The tool's definition is a constant, so the cached tool prefix does not move. Multiple parallel `tool_use` blocks in a single assistant turn fan out concurrently through `router.route_parallel`, each running its own domain-filtered ChromaDB query (3 built-in and 2 company chunks by default, see the RAG section) before the specialist sees the question. The fan-out is failure-isolated per call: a specialist that raises or returns no text becomes an `UNAVAILABLE` tool_result (`router.specialist_unavailable_result`) instead of failing the whole turn, and its area — \"finance\", \"legal\", never the specialist's name — is reported to the user under the web reply, or by the Executive in its reply on other channels, unless another call to it answered that turn. A failed specialist does not count as consulted, so it takes no committee seat. Cancellation (Stop) still propagates. The Executive then synthesizes the returned analyses into one reply.\n\n`remember_fact`, `forget_fact` and `update_company_profile` (`orchestrator/fact_tools.py`) make a correction stick: the first two keep or retire a standing fact in `memory.facts`, and the third changes one field of the company profile. They share `record_decision_outcome`'s gate (the principal on a verified surface, an audited refusal for everyone else) and are in `UNATTENDED_WITHHELD_TOOLS`. `remember_fact` alone also admits a rostered teammate on a verified surface (web signed in, their own Slack or Discord): the fact is stored attributed (`recorded_by_role=\"teammate\"`, rendered `(per <name>)`), and is held as a `proposed` row (`awaiting_approval`) until the principal approves it, unless the principal marked that teammate trusted (\"needs my approval\" off; on by default) — and even then when it would replace the principal's fact. All three are also in `PRIVATE_TURN_WITHHELD_TOOLS`, since what they write or retire is read on everyone's turns. On the principal's own authenticated email (primary address, with Gmail's own `Authentication-Results` reporting `dmarc=pass`, not forwarded) the three tools check the request as usual, then hold it as `awaiting_confirmation`. It applies only when the principal replies to a one-time token emailed to their own address (`integrations.fact_confirmation`); the split into a check step and an `apply_*` step lets chat and the confirmed reply share one apply path. Each write also needs a `source_quote` found in the speaker's own words this turn. The active facts reach every specialist as a `<standing_facts>` user-turn tag: `route_to_specialist` reads them (None → the store, \"\" → none) for chat consults and workflow steps alike, and `BaseAgent.analyze` places the tag next to `<past_decisions>`, never in the cached specialist system prompt.\n\nThe ten entries in `SPECIALIST_REGISTRY` (`openexecutive.orchestrator.router`):\n\n| Key | Agent | Domain |\n| --- | --- | --- |\n| `cso` | `StrategyAgent` | strategy |\n| `cfo` | `FinanceAgent` | finance |\n| `chro` | `HRAgent` | hr |\n| `gc` | `LegalAgent` | legal |\n| `coo` | `OperationsAgent` | operations |\n| `cmo` | `MarketingAgent` | marketing |\n| `cpo` | `ProductAgent` | product |\n| `sales` | `SalesAgent` | sales |\n| `board_comms` | `BoardCommsAgent` | board |\n| `triage` | `TriageAgent` | triage |\n\nModels: specialists default to `claude-sonnet-5-5`; the deep-reasoning tier (CSO, CFO, GC, Board) runs on `DEEP_REASONING_MODEL`, `claude-sonnet-5-5` by default, with adaptive thinking; routing/fast classification uses `claude-haiku-5-5`. Requests are fitted to the model's generation on the way out (`feature_gate.fit_claude_generation`): 5-generation Claude models never receive `temperature` / `top_p` / `top_k`, and Haiku 5.5, which thinks unless told otherwise, gets thinking turned off when the caller asked for none. The Council's **Deep reasoning** toggle is honoured on Claude Opus/Sonnet (direct or via OpenRouter) and on any OpenRouter model whose live catalog entry supports `reasoning` (today all OpenAI, Gemini, DeepSeek and Grok entries; not Llama) — the request translator turns Anthropic adaptive thinking + effort into OpenRouter's `reasoning` parameter. It is a silent no-op on models that can't reason, and disabled for Haiku, which rejects adaptive thinking. A PDF `document` block is translated too: it becomes an OpenAI `file` part, and OpenRouter requests carrying one add the `file-parser` plugin (`native` for models whose catalog entry lists `file` input, else `PDF_OPENROUTER_ENGINE`, default `mistral-ocr`). `FeatureSpec.supports_pdf_input` gates it per model — off for a local server unless `LOCAL_PDF_INPUT` — and a gated-off PDF becomes a text note rather than vanishing. The Council picks a model in two steps: a **Provider** dropdown (Anthropic, OpenAI, Google, Meta, DeepSeek, xAI, Local, …), then a **Model** dropdown limited to that provider, with readable names (\"Claude Opus 5.5\") and the raw id and serving route shown underneath. The Anthropic group combines the direct Claude models (`claude-fable-5-1`, `claude-opus-5-5`, `claude-opus-5`, `claude-sonnet-5-5`, `claude-sonnet-5`, `claude-haiku-5-5`, `claude-haiku-4-5`) with any Claude generation that exists only on OpenRouter, and never lists a model twice. With `OPENROUTER_ENABLED`, the non-Anthropic entries come from OpenRouter's live `/models` catalog (fetched at startup, refreshed in the background, filtered to tool-capable paid text models from an allowlisted vendor set), falling back to a built-in snapshot if the fetch fails. Every Council agent also takes **additional instructions** (`agent_overrides.instructions`): text appended after its prompt by `agents.overrides.append_instructions` inside an `<additional_instructions>` block. `BaseAgent.base_system_prompt()` is the built-in prompt or its full replacement (what the Council's editor shows); `effective_system_prompt()` is that plus the instructions (what every call sends), and the Executive gets the same text through `build_system_blocks(persona_instructions=…)`. Unlike a replaced prompt, instructions leave the agent on its built-in prompt, so prompt improvements keep reaching it. A **Quality** choice above the agents (`agents.presets`) sets every agent's model and deep reasoning at once: **Fast** puts every agent on a small model (`claude-haiku-5-5`, else `claude-haiku-4-5`) with deep reasoning off, **Balanced** returns each agent to its built-in defaults, and **Thorough** puts the Executive and the domain specialists on the largest allowed model (`claude-opus-5-5`, else `claude-opus-5`) with deep reasoning on where the model supports it, leaving triage and the helpers below on their defaults. Presets change quality only; they never cap spend. Each preset resolves against the install's model allowlist at request time and is offered only when one of its models is allowed. Applying one writes ordinary per-agent overrides (model and deep reasoning only), so history, rollback and later per-agent edits keep working; an agent changed on its own afterwards shows as **Custom**. Each Council agent has a **visibility**: `core` for the Executive and the domain specialists (`BaseAgent.visibility = \"core\"`, which a new specialist must declare) and `internal` (the default) for triage and the helper agents below. Thorough moves only core agents. The Council opens in a **simple view**: the Quality choice, the voice picker, and only the core agents, each with its additional instructions and a model badge. **Show all agents** lists the internal ones too, and **Advanced** opens the full page (the browser remembers the choice). Eight Council-only agents are model/prompt-overridable in the UI but NOT callable via `consult_specialist`, because they are deliberately kept out of `SPECIALIST_REGISTRY`: `quality_judge` (committee review), `utility_fast` (gates, titles, decision parsing), `research` (the `executive_research` fan-out), `fixture_generator` (invents demo companies), `engagement_intake` (models a real client from intake material), `onboarding_interviewer` (interviews a new user about their own company and drafts its profile, people and departments), `workflow_designer` (runs the conversational New workflow wizard on `/jobs/new`: searches the available tools, asks clarifying questions, then drafts a validated dynamic-workflow definition for the user to review), and `workflow_actor` (runs a workflow's `action` steps: a bounded tool-use loop limited to the tools the user approved for that step). `fixture_generator`, `engagement_intake` and `onboarding_interviewer` share one shape — a forced tool call producing a whole company, validated for referential integrity, with a single repair retry — but opposite epistemics: the fixture author invents, the other two extract and leave unstated facts empty rather than guessing. When the workspace is in solo mode, `onboarding_interviewer`'s first user turn is prefixed with a short hint (`interview.solo_hint(role_kind)`, one static constant per role kind from the onboarding role step): the draft names only the principal, as principal, and no departments; an owner is asked about the offer, customers, pricing, cash and runway, an in-house executive about their organisation (drafted as the company — their employer), function, remit, manager, team size as context and what they are measured on, an independent about their practice, clients and pricing, and with no role given the role-neutral `SOLO_HINT` asks which they are. The hint rides in the user turn and the tool array is unchanged, so the cached system prompt and tool prefix are the same in both modes.\n\n**Voice.** The Executive answers in one voice persona, substituted into block 0 at `{VOICE_PERSONA}` (`personas/loader.get_active_body`) from the Executive override's `voice_persona_slug`. Three general voices ship in `personas/builtin/`: **Direct** (`default`, the slug used when none is set: answer first, brief and decisive), **Supportive** (a warm coach who explains the why and checks how the plan sits) and **Analytical** (numbers first, options and trade-offs, then one recommendation). They are offered by a shared picker in setup (`/onboard`, after the workspace choice) and in Settings → Executive, which saves through `PATCH /agents/executive`; a built-in voice's frontmatter carries the picker's one-line `description` and `sample` reply. The named-executive voices are marked `legacy: true`: the picker hides them, but they still load, so an install that picked one keeps it, and the Agent Council lists them under \"Legacy voices\". Voice files hold disposition only; the response-length ladder lives in `prompts/executive_persona.py`, and a test keeps numeric length rules out of every voice file. An eval scenario may set `voice_persona` to answer in a voice; it rides on `Session.voice_persona_slug`, which wins over the override for that session only.\n\nOne more structured-output utility agent sits beside `triage`: `openexecutive.agents.alert_review.AlertReviewAgent` (routing model, one forced `emit_alert_reviews` tool call, never raises). The alert review job hands it a batch of open alerts with evidence and gets back, per alert, a verdict (relevant / changed / resolved / stale), a confidence, the evidence it relied on, a one-line note, why-now, an optional deadline, and a recommended move (none / route / nudge / escalate / draft / suggest_workflow / merge / close) with its parameters. Like `triage`, it never mutates state — `alerts.review.apply_verdict` executes the verdicts deterministically within authority, so the model's job is judgment, not action.\n\n**Solo mode.** When the workspace (or the turn's session) is solo — one person using Open Executive for themselves — the Executive works for that one person, the principal, whatever their role: they may run their own business, lead a function inside a larger organisation, or work independently, and the persona reads which from the context instead of assuming it — the role lines under \"Your Principal\" in the org block, when the principal gave their role (workspace settings: kind, title, reports to, remit, measured on). Its persona swaps the team sections (\"You Are a Member of This Executive Team\", \"Choosing Who to Tell\", \"Departments & Org Coordination\") for \"You Work for One Person\", \"The People in Their World\", \"Who Hears From You\" and \"Goals and Areas\". Solo means only the principal uses Open Executive — not that they have no people: their manager, direct reports, peers, clients, investors and board are contacts. The Executive helps the principal work with them (a 1:1 or review prep, an update or budget case for their manager or board, a negotiation) but does not coordinate them, contacts them only when the principal asks and only if they are a contact, and never offers to loop the principal in. Every specialist stays relevant — HR for the principal's own reports, board and investor communications for reporting upward. Goals are grouped by *area* (the tools' `department_slug` is the area's slug). The chat loop withholds `send_department_message`, `send_company_broadcast` and `set_department_head` from its tool list, and a call to one of them anyway gets a JSON error tool_result instead of running. The mode is resolved once per turn and pinned on the session, so the persona, the org block, the tools and every tool handler agree — a switch made mid-turn applies from the next turn. Team mode is unchanged.\n\nEvery specialist *advises* — each returns domain analysis the Executive synthesizes into one voice. None of them act on the company's behalf; the side-effecting tools in the main loop (schedule, people, departments, broadcast, watchlist, workflows) belong to the Executive, not to a specialist.\n\n`recall_history` (Always in the loop, `orchestrator.history_tools`) reads the speaker's own notes, on the same terms as `ghostwrite_email` below: its own registry, offered only on a verified turn in a conversation private to them with their switch on; its `due` flag lists only dated notes, soonest first (see Memory). The ghostwriter below also gets the writer's notes from mail every recipient was on, as a consistency check (`<writer_noted>`).\n\nThe morning reflection, which has no question to look up and no specialist to consult, gets one more read-only tool of its own: `search_knowledge` (`orchestrator.knowledge_tools`), the same knowledge lookup a chat turn makes, for what the company's documents say about a customer, contract, price or policy before it acts.\n\nA fifth, `remind_me` (`orchestrator.reminder_tools`), sets a plain-text reminder sent to the speaker alone when it is due; it opens no mailbox, so it never locks a turn. A sixth, `propose_actions` (`orchestrator.action_card_tools`), does nothing itself: it leaves the speaker a card of up to five actions (a meeting, a message, a new contact) that happen only when they tap Approve, so it stays on after a turn read mail; it is where what an email itself asks for goes, since that is never done on the email's say-so. Four read-only tools beside it read the speaker's own mailbox on the same terms — `search_my_email`, `read_my_email`, `read_my_email_attachment`, `my_email_awaiting_reply` (`orchestrator.mail_read_tools`) — capped per turn, other people's messages and attached files returned as untrusted blocks (see Act as me). One tool writes under someone else's name, and it is fenced in code rather than the prompt. `ghostwrite_email` (**Act as me**, `orchestrator.delegation_tools`) writes an email as the person the Executive is speaking with and saves it as a draft in their own mailbox, Gmail or Outlook (`delegation.gmail.gmail_for` picks the client by the person's credential); chat never sends it (the person sends it themselves, or taps Send on its card). It lives in its own registry, never `_ALL_SKILL_TOOLS`, so reflection, research and workflow toolkits can never carry it. The chat loop adds it only on a turn `delegation.settings.pin_turn_delegation` offered it to — the principal, or a team member once the owner lets them, with it on, verified, in a conversation private to them (the web chat, a Slack or Discord DM, a verified private Telegram chat), on an attended turn — and builds that turn's handler map from both registries, so on any other turn a call to it is an unknown tool. From the round that calls it until the turn ends, the dispatch guard refuses only what opens an outside address, runs a script or workflow, or posts to everyone (`delegation.lockdown`), without changing the offered list; messages and invites to the roster, follow-ups, facts in the person's own words and a contact they name still run, each fenced in its own path. A later turn in the conversation stays private to its owner but holds nothing: the person asking again, having read the reply, is their own request. A contact add the lockdown refuses becomes an approval card in the chat instead (`propose_actions`). Recipients are chosen in code, never by the model (a draft's new email may go to any address, named back to check when it is neither on the roster nor typed), and the text comes from the **ghostwriter** (`delegation.ghostwriter`), a tool-less call with a constant prompt plus the person's \"How I write\" profile. The ghostwriter is deliberately not a Council agent: the Council's edit routes have no principal check. See **Act as Me (Delegation)**. A second tool joins the same way: `recall_history` (`orchestrator.history_tools`, Always in the loop) reads the speaker's own notes of what they said, by email or in chat. It has its own registry and per-turn handler map too, and the loop adds it only when `history_tools.recall_person` finds the speaker verified, in a conversation private to them, on an interactive turn that isn't private to the principal, with their notes switch on. It doesn't need Act as me. After every chat turn the Executive also hands the speaker's own words to `history_chat.schedule_chat_notes`, which takes notes in the background when their switch is on and the surface verified them (see **Memory**).\n\nThe Executive can also launch *any* workflow conversationally: `orchestrator.workflow_run_tools` adds `list_workflows` (the launchable catalog with each workflow's inputs) and `run_workflow` (runs a built-in or custom workflow to completion and returns its artifact, checkpointing approval gates as `awaiting_human`). `run_workflow` runs the weekly review, and in solo mode the morning brief, only for the principal on a verified surface (`Workflow.principal_only_modes`, `create_goal`'s rule); anyone else gets an audited refusal that does not name the rule. A run whose `result` event reports `private_to_principal` (the morning brief on the principal's own verified turn, when it read their private rows) still returns its artifact into that turn, but the stored run keeps `persistence.PRIVATE_RUN_ARTIFACT` instead of the text, since run history is readable by everyone signed in. This makes the `/jobs` catalog reachable from chat — see the Workflows section.\n\nDeliverables are first-class too. `orchestrator.artifact_tools` gives the Executive three tools: `draft_artifact` publishes a memo, web page (`html`), Word document (`docx`), spreadsheet (`xlsx`, from structured `sheets`) or a `link` to something it created in one of the principal's connected apps through the generic MCP gateway (`search_tools` → `call_tool`, any server — no vendor is named), and returns the artifact's id and link; `list_artifacts` and `get_artifact` let it find and reread its own past work (drafts and workflow outputs alike, through the same `artifact_records` read model the API uses). Every artifact belongs to someone: a draft to the person whose turn published it (`artifact_records.current_viewer`; with nobody speaking, the scheduler or the CLI, the principal), and a run someone started by hand to them, while scheduled runs are the team's. The tools, the recall, `message_person`'s link and mail attachments only ever reach the speaker's own (a link is refused when the recipient couldn't open it), not even the principal sees a teammate's, and a speaker who isn't on the People list publishes nothing. A workflow's own session inside someone's turn (the research synthesis) is pinned to that turn's viewer (`Session.documents_viewer`). Revising is `draft_artifact(supersedes=<id>)`, which archives the earlier version. When the principal asks for a deliverable the tool is always used; the \"quiet is the right default\" bar applies only to unprompted drafts. The format enum is a hard-coded sorted tuple so the cached tool schema stays byte-stable. Delivery: `message_person(person_id, text, artifact_id?)` appends the artifact's title and a `UI_BASE_URL/artifacts/<id>` link (through the same outbound guard as any DM; the link needs Open Executive access), and a Gmail send/draft can attach the file itself by `artifact_id` — see Integrations.\n\n**Extensions** (`orchestrator.extensions`). An install can add chat tools and kinds of document without changing this code. `OPENEXECUTIVE_EXTENSIONS` names Python modules. The first turn, or the first Documents read, imports each one and calls its `register()`. A module that fails is logged and skipped. `register_tool` takes an `ExtraTool`: a definition, an async handler, an optional `tool_groups` group (new or existing) and an optional chip. It refuses a name the loop already knows (`executive.builtin_tool_names`). An extension tool is never in `_ALL_SKILL_TOOLS`, so reflection, research and the synthesis never see it. The chat loop adds it per turn, like the Act as me tools, but never on an unattended run, a turn private to the principal or to a speaker who isn't on the People list (mail from a stranger), and only where its own `offered(session)` says. That predicate depends only on the kind of turn, so each kind of turn keeps one cached tool list. Act as me's lockdown refuses it once the turn has read the owner's mail, because it is unclassified and the lockdown fails closed. Its chip goes through `action_chips.summarize_action`, which applies the same no-chip rules for an error or refusal and keeps a link to a path inside the app. `register_collection` adds a kind of document: a drafted artifact tagged `collection:<name>` carries that collection on its record. The Documents page gives each collection its own filter. The collection's `on_change` hears about archive, restore and delete before they happen; if it raises, the change does not happen. `draft_artifact(supersedes=…)` refuses a collection's document. The alert pipeline strips `collection:` tags, so only the extension's own code files a document in its collection. A module that fails partway through `register()` leaves nothing registered. A tool registered with `workflow=True` is also offered to workflow action steps as `oe__<name>` (see Workflows); `read_only=True` marks one that only reads. A step runs with nobody in the turn, so `offered` isn't asked: `extensions.in_workflow_step(name)` is True for that tool, during its call only, and the handler decides what an unattended call may do.\n\nDocuments come in the other way through `read_document` (`orchestrator.document_tools`): it reads a PDF, Word, Excel, CSV, Markdown or text file by `path` — one another tool saved, such as a Gmail attachment (`get_gmail_attachment_content`) or a Drive file (`get_drive_file_download_url`) — or a company document by `filename`, with an optional PDF `pages` range. A scanned PDF is converted (`knowledge.pdf_reader`: local OCR by default; with `PDF_PROVIDER_READING=true`, the deployment's own model through its provider's PDF support, falling back to local OCR) rather than returned empty. Reads are confined to the tool download folders — the same resolve-then-contain check as `oe__read_file` — plus the company docs folder, and it is withheld on turns private to the principal (`PRIVATE_TURN_WITHHELD_TOOLS`): those come from a contact's email, whose own attachments the poller has already read into the turn, and must not steer the Executive into company documents.\n\nThe Executive also backs the **Ask OE** panel's form-fill tier via `orchestrator.form_tools.propose_form_values`: when a chat turn arrives with a `<page_context>` block describing a form on the user's screen, the Executive can propose values for it. The proposal is delivered to the UI as a `form_patch` SSE event and placed into the form as highlighted suggestions — the tool never saves or submits anything; the user reviews and saves manually. The tool is **always** in the tool list (tool definitions sit at the top of the cached prompt prefix, so a conditionally-present tool would invalidate the prompt cache whenever panel and non-panel turns alternate); its description restricts usage to turns that carry a form descriptor.\n\n**Take the lead** (`orchestrator.take_the_lead`) is how the Executive acts on its own, behind one gate. It has two switches, both off until someone turns them on. **As the Executive** is the principal's alone (`PUT /take-the-lead`). With it on, the unattended passes (reflection, research, the scheduler's proactive trigger) keep the acting tools solo mode otherwise withholds from them: booking, starting workflows, and messaging teammates rather than only the principal. `schedule_tools.unattended_toolkit` and the chat loop's unattended path then wrap every acting tool (`ACTING_TOOLS`, plus any MCP tool whose name sends, creates, updates, deletes or shares) in `gated_handlers` / `gated_call_tool`. The principal-only tools (`UNATTENDED_WITHHELD_TOOLS`) stay withheld, and every guard inside a handler still applies, including department approval levels (a booking in a department on Proposes is still a proposal) and the contact-egress rule. New mail or chat (any turn with an `origin_channel`) wakes the existing reflection early (`take_the_lead.wake`): one run waiting at most, five minutes after the first message and no more than every 30 minutes, as a one-off `take_the_lead_wake` scheduled action. Pause holds it like any scheduled work. **As you** is the Handle it for me dial's top step (delegation section).\n\n**The gate** is plain code (`take_the_lead.check`). Six kinds of action wait for a yes first, each with an Ask first switch, all on to start. The switch is the decision ledger's own mode for the class `take_the_lead_action:<kind>` (`propose` asks first, the default; `auto_execute` doesn't). The kinds are money, contracts and legal, hiring and people decisions, deleting/cancelling/sharing, someone new (an address on neither the roster nor the contacts) and big sends (a broadcast or department message, or more than five people). On top of these, the company (set by the principal) and each person (for their own As you) add structured rules: a person or address, a domain, words, or an amount at or above a figure. Added rules always hold, and an error reading them holds the action. The one thing the gate lets straight through is a direct message (`message_person` or a Slack, Telegram or Discord DM) whose only recipient is the principal (`_only_to_principal`): they are who it would wait for, so a card would only ask them to approve a message to themselves. Anything that reaches or acts on anyone else is checked, and a roster that can't be read is not the principal. A held action becomes a `take_the_lead_action` decision instance with a companion `DECISION_ALERT_SOURCE` card on Today, routed to the first teammate holding one of its People approval ranges (`take_the_lead.approval_ranges`, plain code, closest fit first: money to `spend_lt_2k`, `spend_lt_10k` or `spend_gt_10k` by its largest amount, with a higher range able to approve a lower one and no amount read as large, or to `customer_credit` for credit and payment terms; vendor contracts to `vendor_onboarding` then `legal_sign`, other contracts to `legal_sign`; hiring to `hiring_signoff`; anything else about the board or investors to `board_comms`; in Personal as in team mode), or else the principal through the authority gate's fallback (`departments.authority._route_proposal`). Deletes and shares always go to the principal. Only the principal and that approver see or resolve it (`DecisionClassSpec.principal_and_approver`). Approving it claims the row and carries out the exact call once (`take_the_lead.carry_out`: the unwrapped handler, or the gateway's `call_tool`, on an unattended session, with contact egress only when the principal approves). Declining drops it. Everything done or held is a row in `take_the_lead_log`. What it did and what was approved show in Recent activity (`today._build_activity`, kind `took_the_lead`), named but never quoted, since that feed is the team's. The held card's own `decision_resolved` row is left out because its payload quotes the message. `GET /today/activity` also adds the caller's own Handle it sends (kind `sent_as_you`, actor `you`), which are never in the shared feed.\n\n**In training.** The As the Executive switch has three positions: Off, **In training** and On (`take_the_lead.training`). In training, an action the gate would let through still waits, as a `training` hold for the principal alone, unless the principal has allowed that action with that person or thing (`take_the_lead.allowance`: \"Message Priya Nair\" covers a message to her on any channel; \"Book meetings with Priya Nair and Sam Lee\" meetings with exactly them; a workflow by id; a connected tool by name and the file it names, never one that reaches people such as a mail send or a share). The card carries Approve, **Approve + allow**, **Edit** (the message, a meeting's title, times or description, a task and its due date: `take_the_lead.apply_edits`) and Decline. Allowing is the principal's alone and needs a request the API can tie to them, as turning it on does; the six kinds and the added rules hold whatever is allowed. An edit approved with \"Do it like this next time\" is kept as an example (only the edited text, the principal's own words), and `take_the_lead.learned_note` hands the examples and the fact it's in training to the reflection and research passes in their user turn. After `SUGGEST_AFTER` (3) unchanged approvals of the same action in 30 days, Settings suggests allowing it. What it's learned is one list keyed by feature (`take_the_lead_allowed.feature`), so other features that learn from approvals can join it; each keeps its own switch. Act as me is the second (`act_as_me`, per person, `delegation.training`), with a training switch per setting (Replies, Follow-ups, Suggested actions, Drafts): its rows carry the person's id and the setting in their key, are listed and removed by that person alone, count toward their own cap, and never reach `learned_note`.",
+ "markdown": "The Executive (`openexecutive.orchestrator.executive.Executive`) is a single coherent persona that fans out to specialist sub-agents via Anthropic tool use. Each specialist subclasses `openexecutive.agents.base.BaseAgent`, carrying a `name`, `domain`, and `model`, and exposes `analyze(...)`, which wraps the question in user-turn context and calls Claude through the per-call provider. That context is per call and never cached: `<company_stage>` (the stage from the company profile — specialists never see the profile itself), in solo mode `<principal_role>` (what kind of principal the advice is for, in plain words, with their title and remit — `router.principal_role_context`), `<department_memory>`, `<past_decisions>`, `<failure_cases>`, `<relevant_knowledge>` and `<conversation_context>`, outermost first. The specialist's own system prompt is a static domain constant under one `cache_control` block, so it stays byte-identical across companies and profile edits. A chat turn takes the stage from its session's profile (the one the Executive reasons over, and the one an eval scenario injects); workflow steps and the MCP server's `consult_specialist` read it fresh from the profile on disk. The role tag follows the turn's mode: a chat turn pins the role on its session with the mode (`pin_turn_principal_role`: the session's `principal_role` override, else the workspace settings) and sends it in solo and nothing in team, so the org block, every specialist in the turn and any workflow the turn starts see the same role; callers with no chat turn read it fresh (`load_principal_role`: the current session's mode and role, else the workspace's), and a failed read sends no tag rather than failing the consult. The CFO, CSO and sales prompts key their advice on it: for a bootstrapped, self-funded or solo company they drop venture fundraising benchmarks and work from cash, and raise only if the founder wants to.\n\nRouting uses one `consult_specialist` tool. Beside it the Executive carries two skill tools, `search_skills` and `load_skill` (skills are shown to users as **playbooks**, and every playbook belongs to a workflow: one that follows it, or its own Quick workflow). Chat finds and reads playbooks but never creates or changes one: work worth repeating is offered as a workflow through `draft_workflow`, with the method written into its steps, because workflows read playbooks at run time (approved scheduled ones included) and this loop also runs on inbound email and chat, so only a person can change what a workflow follows. A hidden built-in is invisible to search and load; each `search_skills` hit lists the `workflows` that follow it, Quick ones included, and the persona routes \"the finished deliverable\" to that workflow and quick or partial asks to the playbook inline; Tool calls made while proactive work is tagged (scheduler dispatch, reflection, research, alert review) feed the Attunement outcome ledger. Goals have three tools: `list_department_goals`, `update_department_goal` (progress the user reports) and `create_goal` (a new goal the principal states, filed under an area by slug or title and creating the area when none matches — see Org). `create_goal` runs only for the principal on a verified surface, and no unattended run is offered it: the tools in `schedule_tools.UNATTENDED_WITHHELD_TOOLS` are dropped from reflection's and research's toolkits by `unattended_toolkit`, and from the chat loop when its Session is marked `unattended` — the scheduler's PROACTIVE TRIGGER run, whose prompt quotes stored intent text; a call the model emits anyway gets an error tool_result. `record_decision_outcome` records how a past decision turned out — the weekly review lists decisions older than 30 days with no outcome as `[decision N]` — writing `decisions.outcome` (`episodic.update_decision`) and auditing the rationale; it has `create_goal`'s gate (the principal on a verified surface only; everyone else gets an audited refusal) and is in `UNATTENDED_WITHHELD_TOOLS` too. A missing or unknown id returns the decisions still waiting on an outcome, so the model can retry with the right one. Its schema is static; the solo persona has one line on it, and team mode learns it from the tool description, so the team persona is byte-identical. Attunement adds `list_open_loops` (what a person owes — everyone's for the principal only in a private conversation, otherwise the caller's own) and `close_open_loop`, which only closes a loop when the turn's resolved speaker is the principal or the loop's owner — an unrostered sender, a signed-in web user not on the roster, and unattended sessions are refused. `assign_open_loop` opens one on request (\"ask Ben to send me the Q3 numbers by Friday\") for the principal or a rostered teammate on a verified surface — never an email, a turn private to the principal or an unattended run, which are also withheld the tool — and only when `task` appears verbatim in what the speaker typed this turn, at most 5 per turn; that turn's extraction pass then opens no second loop for the same owner. The roster write tools (`upsert_person`, `archive_person`, `set_department_head`) are the principal's alone, and only on a surface that verified the speaker — the web chat, the principal's own Slack or Discord, or a private Telegram chat when `TELEGRAM_WEBHOOK_SECRET` is set to a value Telegram can send (Telegram identifies the chat rather than the sender, and without that secret its webhook accepts anyone's update). A teammate, an email (even from the principal's address), Google Chat, the CLI, the MCP server and unattended runs get `status: refused` (audited, and no action chip), because a team member's People row is also the web sign-in and outbound-email allow-list. `upsert_person` takes `kind` (`team` / `contact`; a new person defaults to `contact` in solo mode, an update keeps the current kind) and `email_aliases` (other addresses the person writes from — the full list, replacing the current one; they match mail and may be emailed but never sign in, and one already on someone else is refused), and after every write any pending roster request whose sender now matches someone is closed and its held messages replayed. `resolve_roster_request` answers a **roster request** — someone not on the People list who wrote in and is waiting (see Org): `approve` (a new person; `kind` defaults to team for a company-domain sender, else contact, when the principal did not say), `link` (someone already listed, by `person_id`; `replace_channel_id` only when the principal said so) or `decline`. It has the roster tools' gate (the principal on a verified surface; audited refusal otherwise), is in `UNATTENDED_WITHHELD_TOOLS`, and answers only an id in `Session.trusted_roster_request_ids` — the ids the server printed that turn in a `<roster_requests>` block (`briefing.context.format_roster_requests_for_prompt`, up to 10 pending requests), which `render_and_trust` appends to the `<briefing>` digest on the principal's own verified turn only and records on the session in the same step, so an id the model invents or reads out of a message adds nobody. Each line is server-derived — id, channel, address or account id, the sender's name sanitised and marked unverified, a company-domain hint, the message count — and never what they wrote. The block rides in the **user turn**, like the digest, so the cached system blocks are unchanged; the tool definition is static, and the new tool only extends the sorted tool list. The web chat route resets the trusted set every turn, and a resolved id is removed from it. and `list_people` returns both kinds with `kind` — contacts only on the principal's own verified turn (they are private to the principal; anyone else gets the team as if no contact existed) — while `lookup_person` searches the team only. `ask_about_person` reads a person's peer memory fully only on their own verified turn; anyone else's answers only to a verified principal or teammate, and only when that person turned on \"Share my work style with the team\", framed to how they work and never their view of a colleague, and otherwise returns the same \"hasn't chosen to share\" result whether or not anything is known. A teammate's own mail with the Executive is theirs the same way: every mailbox call the model makes through the MCP gateway (Gmail or Outlook, by `call_tool`, from a script, or a workflow step run from chat) passes through `orchestrator.teammate_mail`. A search excludes every other teammate's address server-side and is filtered again entry by entry; any other call that names a message or thread (read, attachment, reply, forward, label) must use an id a filtered search returned to the same reader in this conversation (`Session.mail_ids_seen`) or the inbound email the turn answers; reads are filtered message by message, so a message naming another teammate is left out unless the reader is in its From / To / Cc; a reply that would quote such a thread, or an Outlook forward, reply or attachment read of such a message, is refused; and a conversation in which a teammate other than the principal was shown their own mail becomes theirs alone, as Act as me's do. Outside mail and the reader's own are untouched; an email turn reads as its sender only when Gmail authenticated it, and any other unverified reader sees no teammate's mail, a background run the model drives reads as the principal and a workflow step outside any session as nobody, an Outlook inbox rule that forwards or names another teammate is refused, and the poller and the brief, which read the mailbox themselves, are unaffected. Their own chats are theirs too: every audit row of a teammate's verified turn is theirs (`people_tools.teammate_turn_owner`, see Audit; the principal sees the actions, never the words), and a teammate's web chat or DM opens for them alone (`chat._session_access`), the principal included. The same verified-principal rule, as `people_tools.contacts_reachable_now()`, decides whether a turn may reach a **contact** at all: `message_person`, the raw DM tools, `create_calendar_event` / `create_instant_meeting` attendees and the MCP gateway's Gmail / Calendar / Drive / OneDrive gates treat a contact exactly like an unknown person on every other turn, so no refusal reveals that one exists. **The untrusted-content policy** (`orchestrator.content_trust`) is one rule for text the principal did not write — inbound email, chat, attached files, watched pages — answering three questions for every surface. *Who is speaking*: the principal only when the surface proved it (`principal_speaking`: the web chat as the principal — never a signed-in email on nobody's People entry —, the CLI, their own Slack/Discord or a secret-verified Telegram chat, or their own address with Gmail's `dmarc=pass`); never a teammate, a stranger, Google Chat, the MCP server or an unattended run. *What may become memory*: episodic extraction runs only then, and only on the words outside `<untrusted_content>` blocks; peer memory and open loops are recorded under the speaker's own name, never as the principal's. *What may trigger tools*: every write on the principal's authority already refuses others in its handler; the tools with no such check that change the install for every later turn — `content_trust.PRINCIPAL_ONLY_TOOLS`: `load_mcp_server`, `list_saved_tools`, `run_python_job` and `run_script` — are not offered unless the principal is speaking on an interactive surface (not email, even authenticated), and a call the model emits anyway gets an error tool_result and a `refused: not_principal` audit row. So an inbound email, a teammate's turn, Google Chat and the scheduler's proactive trigger each get a stable tool list of their own without it. *How it is labelled*: outside text goes into the user turn inside `<untrusted_content source=… author=… author_verified=…>` with a fixed notice that it carries no authority; compatibility forms are folded (NFKC), control characters go, and the tag's name is renamed wherever the text carries it — any case, spacing or entity-escaped form — so the text cannot end its block and speak outside it.\n\nA turn about the principal's private mail (`Session.private_to_principal`) reaches the principal only. It is neither offered nor allowed to run `schedule_tools.PRIVATE_TURN_WITHHELD_TOOLS` — everything that reaches someone else, publishes where the team reads, or starts work outside the turn: the broadcasts, meeting bookings and cancellations, `run_workflow` and research, follow-ups and workflow nudges, watches, artifacts, goal updates, saved workflows, skill drafts, and `load_mcp_server` (it connects to any URL, which can carry the turn's content). Through `search_tools` and `call_tool` it reaches only `schedule_tools.PRIVATE_TURN_MCP_TOOLS`: Google Workspace reads (mail, calendar, and Drive, Docs and Sheets search and file reads, with their Microsoft 365 twins; OneDrive's `download-bytes` only for a file's content) and the Gmail send and draft the gateway narrows to the principal. A search shows it nothing else, and a call to anything else (Google Chat, a Docs or Drive write, `manage_event`, another server) is refused (`schedule_tools.private_turn_withholds`). A call the model emits anyway, MCP tools included, gets an error tool_result and writes an audit row, and every audit row the turn writes is the principal's alone. On every turn the Executive knows what is connected without searching: block 0's *Connected Systems* section names the connected Google services, other MCP servers, the messaging channels, Notion sync, Confluence sync (when on) and Act as me, and pins the Google tool names it calls directly with `call_tool`. Both the service list and the pinned names are narrowed to what workspace-mcp actually loaded (`WORKSPACE_MCP_TOOLS`, `WORKSPACE_MCP_TOOL_TIER`), so a service left out is not named and its tools are not pinned — being told a tool is already available when it is not in the index costs a failed call the model cannot explain to itself. `search_tools` is for anything not listed. When work repeats over many items, the Executive can also call `run_script` (`workflows/step_script.py`, offered when the gateway runs, unless `CHAT_SCRIPTS=false`, and only on the principal's own turns — `PRINCIPAL_ONLY_TOOLS`, withheld on an unattended pass too, because it spends the turn's whole call budget on code the model just wrote): one short Python script, run in a sandboxed Monty worker with no files, network or environment, whose only reach is the gateway tools `call_tool` could reach (those `search_tools` has returned; the gateway keeps that set for its own process-wide session, not per conversation). Each call it makes is a `call_tool` in every way that matters — the gateway's discovery, deny-list and recipient gates, Take the lead's gate on an unattended run, and the same action chip and `tool_invocation` row (marked `via: run_script`, `ok: false` when it raised), as each call happens; the script itself gets a `kind: script` row with its source and result. Its wall clock is half the turn's deadline, so a long script ends with the list of calls it made rather than being cut off with the turn, and a stopped turn closes it and its worker. A turn private to the principal is not offered it, and the Act as me lockdown withholds it whole. Besides gateway tools, a chat script may call a fixed list of the Executive's own tools (`step_script.CHAT_OWN_TOOLS`: reads, alerts, open loops, goals, the watch list, follow-ups, `message_person`, `upsert_person` — nothing that broadcasts, books calendars, edits memory or skills, or starts workflows or research), each only when the turn is offered it and through the turn's own handler (Take the lead's gate included), with the same chip and a `kind: skill`, `via: run_script` audit row. A direct call needs a `tool_use` each, so a turn's scripts share a budget instead: `CHAT_SCRIPT_MAX_CALLS` (default 200) calls in all, and per-turn caps on own tools that could otherwise act in bulk (`CHAT_OWN_TOOL_CAPS`: `message_person` 10 — messaging the whole roster would be a broadcast — `upsert_person` 25, `create_alert` 50); a refused call tells the model to say what is left. A call stopped mid-flight by the script's clock or the turn still gets a `cancelled` audit row, since it may have run. Because every round re-reads the conversation, the loop also nudges at the moment of choice: after a round whose results include a list of five or more items (`step_script.lists_many`), with `run_script` on offer and not used that round, it adds the constant `FANOUT_HINT` as a text block in that round's user message (never a cached block), once a turn, pointing the per-item work at one built tool. For library work on files (PDFs, spreadsheets, charts, images) the chat also has `run_python_job` (see Workflows → Python jobs): Python with a fixed library set in a Deno-isolated WebAssembly sandbox, fed Knowledge files by name, the message's own attachments as sent, or text it already holds, and able to keep a job that worked as a Custom tool, with result files linked for the principal to download; it is offered only on the principal's own turns and only where the sandbox is installed (not offered otherwise, so the tool prefix stays fixed per process). Agent Activity shows each run as one \"Built tool\" card (debug event `script_run`: ok, the calls it made, how many, how long, and the kept or reused tool's name). The persona's tool-building section is present only when chat scripts can actually run (the setting and Monty installed). `TOOL_BUILDING_ADDENDUM` in the cached persona tells the Executive when to build a tool — a job that looks something up and then checks or acts on each result, paged results, more than about ten items, or an answer that needs three or more reads (find, open, read; the tool returns only what answers the question, not whole documents, and only what is about to be edited or quoted at length is read in full), while a handful of items already named go as direct calls in one turn — and never to say it has none, and to talk about it as building a tool or doing it in one go. On the principal's own verified, interactive turns, a script that works can be saved as a named tool (`save_as`; elsewhere the script runs but is not saved) and run again with `run_script(tool=…, inputs=…)`; `list_saved_tools` (a constant definition, in `PRINCIPAL_ONLY_TOOLS` with `load_mcp_server`, so offered only while the principal is speaking on a verified, interactive surface; `run_script(tool=…)` is refused on any other turn) lists them, and each call a saved tool makes is a gateway call like any other (see Workflows → Saved tools). The tool's definition is a constant, so the cached tool prefix does not move. Multiple parallel `tool_use` blocks in a single assistant turn fan out concurrently through `router.route_parallel`, each running its own domain-filtered ChromaDB query (3 built-in and 2 company chunks by default, see the RAG section) before the specialist sees the question. The fan-out is failure-isolated per call: a specialist that raises or returns no text becomes an `UNAVAILABLE` tool_result (`router.specialist_unavailable_result`) instead of failing the whole turn, and its area — \"finance\", \"legal\", never the specialist's name — is reported to the user under the web reply, or by the Executive in its reply on other channels, unless another call to it answered that turn. A failed specialist does not count as consulted, so it takes no committee seat. Cancellation (Stop) still propagates. The Executive then synthesizes the returned analyses into one reply.\n\n`remember_fact`, `forget_fact` and `update_company_profile` (`orchestrator/fact_tools.py`) make a correction stick: the first two keep or retire a standing fact in `memory.facts`, and the third changes one field of the company profile. They share `record_decision_outcome`'s gate (the principal on a verified surface, an audited refusal for everyone else) and are in `UNATTENDED_WITHHELD_TOOLS`. `remember_fact` alone also admits a rostered teammate on a verified surface (web signed in, their own Slack or Discord): the fact is stored attributed (`recorded_by_role=\"teammate\"`, rendered `(per <name>)`), and is held as a `proposed` row (`awaiting_approval`) until the principal approves it, unless the principal marked that teammate trusted (\"needs my approval\" off; on by default) — and even then when it would replace the principal's fact. All three are also in `PRIVATE_TURN_WITHHELD_TOOLS`, since what they write or retire is read on everyone's turns. On the principal's own authenticated email (primary address, with Gmail's own `Authentication-Results` reporting `dmarc=pass`, not forwarded) the three tools check the request as usual, then hold it as `awaiting_confirmation`. It applies only when the principal replies to a one-time token emailed to their own address (`integrations.fact_confirmation`); the split into a check step and an `apply_*` step lets chat and the confirmed reply share one apply path. Each write also needs a `source_quote` found in the speaker's own words this turn. The active facts reach every specialist as a `<standing_facts>` user-turn tag: `route_to_specialist` reads them (None → the store, \"\" → none) for chat consults and workflow steps alike, and `BaseAgent.analyze` places the tag next to `<past_decisions>`, never in the cached specialist system prompt.\n\nThe ten entries in `SPECIALIST_REGISTRY` (`openexecutive.orchestrator.router`):\n\n| Key | Agent | Domain |\n| --- | --- | --- |\n| `cso` | `StrategyAgent` | strategy |\n| `cfo` | `FinanceAgent` | finance |\n| `chro` | `HRAgent` | hr |\n| `gc` | `LegalAgent` | legal |\n| `coo` | `OperationsAgent` | operations |\n| `cmo` | `MarketingAgent` | marketing |\n| `cpo` | `ProductAgent` | product |\n| `sales` | `SalesAgent` | sales |\n| `board_comms` | `BoardCommsAgent` | board |\n| `triage` | `TriageAgent` | triage |\n\nModels: specialists default to `claude-sonnet-5-5`; the deep-reasoning tier (CSO, CFO, GC, Board) runs on `DEEP_REASONING_MODEL`, `claude-sonnet-5-5` by default, with adaptive thinking; routing/fast classification uses `claude-haiku-5-5`. Requests are fitted to the model's generation on the way out (`feature_gate.fit_claude_generation`): 5-generation Claude models never receive `temperature` / `top_p` / `top_k`, and Haiku 5.5, which thinks unless told otherwise, gets thinking turned off when the caller asked for none. The Council's **Deep reasoning** toggle is honoured on Claude Opus/Sonnet (direct or via OpenRouter) and on any OpenRouter model whose live catalog entry supports `reasoning` (today all OpenAI, Gemini, DeepSeek and Grok entries; not Llama) — the request translator turns Anthropic adaptive thinking + effort into OpenRouter's `reasoning` parameter. It is a silent no-op on models that can't reason, and disabled for Haiku, which rejects adaptive thinking. A PDF `document` block is translated too: it becomes an OpenAI `file` part, and OpenRouter requests carrying one add the `file-parser` plugin (`native` for models whose catalog entry lists `file` input, else `PDF_OPENROUTER_ENGINE`, default `mistral-ocr`). `FeatureSpec.supports_pdf_input` gates it per model — off for a local server unless `LOCAL_PDF_INPUT` — and a gated-off PDF becomes a text note rather than vanishing. The Council picks a model in two steps: a **Provider** dropdown (Anthropic, OpenAI, Google, Meta, DeepSeek, xAI, Local, …), then a **Model** dropdown limited to that provider, with readable names (\"Claude Opus 5.5\") and the raw id and serving route shown underneath. The Anthropic group combines the direct Claude models (`claude-fable-5-1`, `claude-opus-5-5`, `claude-opus-5`, `claude-sonnet-5-5`, `claude-sonnet-5`, `claude-haiku-5-5`, `claude-haiku-4-5`) with any Claude generation that exists only on OpenRouter, and never lists a model twice. With `OPENROUTER_ENABLED`, the non-Anthropic entries come from OpenRouter's live `/models` catalog (fetched at startup, refreshed in the background, filtered to tool-capable paid text models from an allowlisted vendor set), falling back to a built-in snapshot if the fetch fails. Every Council agent also takes **additional instructions** (`agent_overrides.instructions`): text appended after its prompt by `agents.overrides.append_instructions` inside an `<additional_instructions>` block. `BaseAgent.base_system_prompt()` is the built-in prompt or its full replacement (what the Council's editor shows); `effective_system_prompt()` is that plus the instructions (what every call sends), and the Executive gets the same text through `build_system_blocks(persona_instructions=…)`. Unlike a replaced prompt, instructions leave the agent on its built-in prompt, so prompt improvements keep reaching it. A **Quality** choice above the agents (`agents.presets`) sets every agent's model and deep reasoning at once: **Fast** puts every agent on a small model (`claude-haiku-5-5`, else `claude-haiku-4-5`) with deep reasoning off, **Balanced** returns each agent to its built-in defaults, and **Thorough** puts the Executive and the domain specialists on the largest allowed model (`claude-opus-5-5`, else `claude-opus-5`) with deep reasoning on where the model supports it, leaving triage and the helpers below on their defaults. Presets change quality only; they never cap spend. Each preset resolves against the install's model allowlist at request time and is offered only when one of its models is allowed. Applying one writes ordinary per-agent overrides (model and deep reasoning only), so history, rollback and later per-agent edits keep working; an agent changed on its own afterwards shows as **Custom**. Each Council agent has a **visibility**: `core` for the Executive and the domain specialists (`BaseAgent.visibility = \"core\"`, which a new specialist must declare) and `internal` (the default) for triage and the helper agents below. Thorough moves only core agents. The Council opens in a **simple view**: the Quality choice, the voice picker, and only the core agents, each with its additional instructions and a model badge. **Show all agents** lists the internal ones too, and **Advanced** opens the full page (the browser remembers the choice). Eight Council-only agents are model/prompt-overridable in the UI but NOT callable via `consult_specialist`, because they are deliberately kept out of `SPECIALIST_REGISTRY`: `quality_judge` (committee review), `utility_fast` (gates, titles, decision parsing), `research` (the `executive_research` fan-out), `fixture_generator` (invents demo companies), `engagement_intake` (models a real client from intake material), `onboarding_interviewer` (interviews a new user about their own company and drafts its profile, people and departments), `workflow_designer` (runs the conversational New workflow wizard on `/jobs/new`: searches the available tools, asks clarifying questions, then drafts a validated dynamic-workflow definition for the user to review), and `workflow_actor` (runs a workflow's `action` steps: a bounded tool-use loop limited to the tools the user approved for that step). `fixture_generator`, `engagement_intake` and `onboarding_interviewer` share one shape — a forced tool call producing a whole company, validated for referential integrity, with a single repair retry — but opposite epistemics: the fixture author invents, the other two extract and leave unstated facts empty rather than guessing. When the workspace is in solo mode, `onboarding_interviewer`'s first user turn is prefixed with a short hint (`interview.solo_hint(role_kind)`, one static constant per role kind from the onboarding role step): the draft names only the principal, as principal, and no departments; an owner is asked about the offer, customers, pricing, cash and runway, an in-house executive about their organisation (drafted as the company — their employer), function, remit, manager, team size as context and what they are measured on, an independent about their practice, clients and pricing, and with no role given the role-neutral `SOLO_HINT` asks which they are. The hint rides in the user turn and the tool array is unchanged, so the cached system prompt and tool prefix are the same in both modes.\n\n**Voice.** The Executive answers in one voice persona, substituted into block 0 at `{VOICE_PERSONA}` (`personas/loader.get_active_body`) from the Executive override's `voice_persona_slug`. Three general voices ship in `personas/builtin/`: **Direct** (`default`, the slug used when none is set: answer first, brief and decisive), **Supportive** (a warm coach who explains the why and checks how the plan sits) and **Analytical** (numbers first, options and trade-offs, then one recommendation). They are offered by a shared picker in setup (`/onboard`, after the workspace choice) and in Settings → Executive, which saves through `PATCH /agents/executive`; a built-in voice's frontmatter carries the picker's one-line `description` and `sample` reply. The named-executive voices are marked `legacy: true`: the picker hides them, but they still load, so an install that picked one keeps it, and the Agent Council lists them under \"Legacy voices\". Voice files hold disposition only; the response-length ladder lives in `prompts/executive_persona.py`, and a test keeps numeric length rules out of every voice file. An eval scenario may set `voice_persona` to answer in a voice; it rides on `Session.voice_persona_slug`, which wins over the override for that session only.\n\nOne more structured-output utility agent sits beside `triage`: `openexecutive.agents.alert_review.AlertReviewAgent` (routing model, one forced `emit_alert_reviews` tool call, never raises). The alert review job hands it a batch of open alerts with evidence and gets back, per alert, a verdict (relevant / changed / resolved / stale), a confidence, the evidence it relied on, a one-line note, why-now, an optional deadline, and a recommended move (none / route / nudge / escalate / draft / suggest_workflow / merge / close) with its parameters. Like `triage`, it never mutates state — `alerts.review.apply_verdict` executes the verdicts deterministically within authority, so the model's job is judgment, not action.\n\n**Solo mode.** When the workspace (or the turn's session) is solo — one person using Open Executive for themselves — the Executive works for that one person, the principal, whatever their role: they may run their own business, lead a function inside a larger organisation, or work independently, and the persona reads which from the context instead of assuming it — the role lines under \"Your Principal\" in the org block, when the principal gave their role (workspace settings: kind, title, reports to, remit, measured on). Its persona swaps the team sections (\"You Are a Member of This Executive Team\", \"Choosing Who to Tell\", \"Departments & Org Coordination\") for \"You Work for One Person\", \"The People in Their World\", \"Who Hears From You\" and \"Goals and Areas\". Solo means only the principal uses Open Executive — not that they have no people: their manager, direct reports, peers, clients, investors and board are contacts. The Executive helps the principal work with them (a 1:1 or review prep, an update or budget case for their manager or board, a negotiation) but does not coordinate them, contacts them only when the principal asks and only if they are a contact, and never offers to loop the principal in. Every specialist stays relevant — HR for the principal's own reports, board and investor communications for reporting upward. Goals are grouped by *area* (the tools' `department_slug` is the area's slug). The chat loop withholds `send_department_message`, `send_company_broadcast` and `set_department_head` from its tool list, and a call to one of them anyway gets a JSON error tool_result instead of running. The mode is resolved once per turn and pinned on the session, so the persona, the org block, the tools and every tool handler agree — a switch made mid-turn applies from the next turn. Team mode is unchanged.\n\nEvery specialist *advises* — each returns domain analysis the Executive synthesizes into one voice. None of them act on the company's behalf; the side-effecting tools in the main loop (schedule, people, departments, broadcast, watchlist, workflows) belong to the Executive, not to a specialist.\n\n`recall_history` (Always in the loop, `orchestrator.history_tools`) reads the speaker's own notes, on the same terms as `ghostwrite_email` below: its own registry, offered only on a verified turn in a conversation private to them with their switch on; its `due` flag lists only dated notes, soonest first (see Memory). The ghostwriter below also gets the writer's notes from mail every recipient was on, as a consistency check (`<writer_noted>`).\n\nThe morning reflection, which has no question to look up and no specialist to consult, gets one more read-only tool of its own: `search_knowledge` (`orchestrator.knowledge_tools`), the same knowledge lookup a chat turn makes, for what the company's documents say about a customer, contract, price or policy before it acts.\n\nA fifth, `remind_me` (`orchestrator.reminder_tools`), sets a plain-text reminder sent to the speaker alone when it is due; it opens no mailbox, so it never locks a turn. A sixth, `propose_actions` (`orchestrator.action_card_tools`), does nothing itself: it leaves the speaker a card of up to five actions (a meeting, a message, a new contact) that happen only when they tap Approve, so it stays on after a turn read mail; it is where what an email itself asks for goes, since that is never done on the email's say-so. Four read-only tools beside it read the speaker's own mailbox on the same terms — `search_my_email`, `read_my_email`, `read_my_email_attachment`, `my_email_awaiting_reply` (`orchestrator.mail_read_tools`) — capped per turn, other people's messages and attached files returned as untrusted blocks (see Act as me). One tool writes under someone else's name, and it is fenced in code rather than the prompt. `ghostwrite_email` (**Act as me**, `orchestrator.delegation_tools`) writes an email as the person the Executive is speaking with and saves it as a draft in their own mailbox, Gmail or Outlook (`delegation.gmail.gmail_for` picks the client by the person's credential); chat never sends it (the person sends it themselves, or taps Send on its card). It lives in its own registry, never `_ALL_SKILL_TOOLS`, so reflection, research and workflow toolkits can never carry it. The chat loop adds it only on a turn `delegation.settings.pin_turn_delegation` offered it to — the principal, or a team member once the owner lets them, with it on, verified, in a conversation private to them (the web chat, a Slack or Discord DM, a verified private Telegram chat), on an attended turn — and builds that turn's handler map from both registries, so on any other turn a call to it is an unknown tool. From the round that calls it until the turn ends, the dispatch guard refuses only what opens an outside address, runs a script or workflow, or posts to everyone (`delegation.lockdown`), without changing the offered list; messages and invites to the roster, follow-ups, facts in the person's own words and a contact they name still run, each fenced in its own path. A later turn in the conversation stays private to its owner but holds nothing: the person asking again, having read the reply, is their own request. A contact add the lockdown refuses becomes an approval card in the chat instead (`propose_actions`). Recipients are chosen in code, never by the model (a draft's new email may go to any address, named back to check when it is neither on the roster nor typed), and the text comes from the **ghostwriter** (`delegation.ghostwriter`), a tool-less call with a constant prompt plus the person's \"How I write\" profile. The ghostwriter is deliberately not a Council agent: the Council's edit routes have no principal check. See **Act as Me (Delegation)**. A second tool joins the same way: `recall_history` (`orchestrator.history_tools`, Always in the loop) reads the speaker's own notes of what they said, by email or in chat. It has its own registry and per-turn handler map too, and the loop adds it only when `history_tools.recall_person` finds the speaker verified, in a conversation private to them, on an interactive turn that isn't private to the principal, with their notes switch on. It doesn't need Act as me. After every chat turn the Executive also hands the speaker's own words to `history_chat.schedule_chat_notes`, which takes notes in the background when their switch is on and the surface verified them (see **Memory**).\n\nThe Executive can also launch *any* workflow conversationally: `orchestrator.workflow_run_tools` adds `list_workflows` (the launchable catalog with each workflow's inputs) and `run_workflow` (runs a built-in or custom workflow to completion and returns its artifact, checkpointing approval gates as `awaiting_human`). `run_workflow` runs the weekly review, and in solo mode the morning brief, only for the principal on a verified surface (`Workflow.principal_only_modes`, `create_goal`'s rule); anyone else gets an audited refusal that does not name the rule. A run whose `result` event reports `private_to_principal` (the morning brief on the principal's own verified turn, when it read their private rows) still returns its artifact into that turn, but the stored run keeps `persistence.PRIVATE_RUN_ARTIFACT` instead of the text, since run history is readable by everyone signed in. This makes the `/jobs` catalog reachable from chat — see the Workflows section.\n\nDeliverables are first-class too. `orchestrator.artifact_tools` gives the Executive three tools: `draft_artifact` publishes a memo, web page (`html`), Word document (`docx`), spreadsheet (`xlsx`, from structured `sheets`) or a `link` to something it created in one of the principal's connected apps through the generic MCP gateway (`search_tools` → `call_tool`, any server — no vendor is named), and returns the artifact's id and link; `list_artifacts` and `get_artifact` let it find and reread its own past work (drafts and workflow outputs alike, through the same `artifact_records` read model the API uses). Every artifact belongs to someone: a draft to the person whose turn published it (`artifact_records.current_viewer`; with nobody speaking, the scheduler or the CLI, the principal), and a run someone started by hand to them, while scheduled runs are the team's. The tools, the recall, `message_person`'s link and mail attachments only ever reach the speaker's own (a link is refused when the recipient couldn't open it), not even the principal sees a teammate's, and a speaker who isn't on the People list publishes nothing. A workflow's own session inside someone's turn (the research synthesis) is pinned to that turn's viewer (`Session.documents_viewer`). Revising is `draft_artifact(supersedes=<id>)`, which archives the earlier version. When the principal asks for a deliverable the tool is always used; the \"quiet is the right default\" bar applies only to unprompted drafts. The format enum is a hard-coded sorted tuple so the cached tool schema stays byte-stable. Delivery: `message_person(person_id, text, artifact_id?)` appends the artifact's title and a `UI_BASE_URL/artifacts/<id>` link (through the same outbound guard as any DM; the link needs Open Executive access), and a Gmail send/draft can attach the file itself by `artifact_id` — see Integrations.\n\n**Extensions** (`orchestrator.extensions`). An install can add chat tools and kinds of document without changing this code. `OPENEXECUTIVE_EXTENSIONS` names Python modules. The first turn, or the first Documents read, imports each one and calls its `register()`. A module that fails is logged and skipped. `register_tool` takes an `ExtraTool`: a definition, an async handler, an optional `tool_groups` group (new or existing) and an optional chip. It refuses a name the loop already knows (`executive.builtin_tool_names`). An extension tool is never in `_ALL_SKILL_TOOLS`, so reflection, research and the synthesis never see it. The chat loop adds it per turn, like the Act as me tools, but never on an unattended run, a turn private to the principal or to a speaker who isn't on the People list (mail from a stranger), and only where its own `offered(session)` says. That predicate depends only on the kind of turn, so each kind of turn keeps one cached tool list. Act as me's lockdown refuses it once the turn has read the owner's mail, because it is unclassified and the lockdown fails closed. Its chip goes through `action_chips.summarize_action`, which applies the same no-chip rules for an error or refusal and keeps a link to a path inside the app. `register_collection` adds a kind of document: a drafted artifact tagged `collection:<name>` carries that collection on its record. The Documents page gives each collection its own filter. The collection's `on_change` hears about archive, restore and delete before they happen; if it raises, the change does not happen. `draft_artifact(supersedes=…)` refuses a collection's document. The alert pipeline strips `collection:` tags, so only the extension's own code files a document in its collection. A module that fails partway through `register()` leaves nothing registered. A tool registered with `workflow=True` is also offered to workflow action steps as `oe__<name>` (see Workflows); `read_only=True` marks one that only reads. A step runs with nobody in the turn, so `offered` isn't asked: `extensions.in_workflow_step(name)` is True for that tool, during its call only, and the handler decides what an unattended call may do.\n\nDocuments come in the other way through `read_document` (`orchestrator.document_tools`): it reads a PDF, Word, Excel, CSV, Markdown or text file by `path` — one another tool saved, such as a Gmail attachment (`get_gmail_attachment_content`) or a Drive file (`get_drive_file_download_url`) — or a company document by `filename`, with an optional PDF `pages` range. A scanned PDF is converted (`knowledge.pdf_reader`: local OCR by default; with `PDF_PROVIDER_READING=true`, the deployment's own model through its provider's PDF support, falling back to local OCR) rather than returned empty. Reads are confined to the tool download folders — the same resolve-then-contain check as `oe__read_file` — plus the company docs folder, and it is withheld on turns private to the principal (`PRIVATE_TURN_WITHHELD_TOOLS`): those come from a contact's email, whose own attachments the poller has already read into the turn, and must not steer the Executive into company documents.\n\nThe Executive also backs the **Ask OE** panel's form-fill tier via `orchestrator.form_tools.propose_form_values`: when a chat turn arrives with a `<page_context>` block describing a form on the user's screen, the Executive can propose values for it. The proposal is delivered to the UI as a `form_patch` SSE event and placed into the form as highlighted suggestions — the tool never saves or submits anything; the user reviews and saves manually. The tool is **always** in the tool list (tool definitions sit at the top of the cached prompt prefix, so a conditionally-present tool would invalidate the prompt cache whenever panel and non-panel turns alternate); its description restricts usage to turns that carry a form descriptor.\n\n**Take the lead** (`orchestrator.take_the_lead`) is how the Executive acts on its own, behind one gate. It has two switches, both off until someone turns them on. **As the Executive** is the principal's alone (`PUT /take-the-lead`). With it on, the unattended passes (reflection, research, the scheduler's proactive trigger) keep the acting tools solo mode otherwise withholds from them: booking, starting workflows, and messaging teammates rather than only the principal. `schedule_tools.unattended_toolkit` and the chat loop's unattended path then wrap every acting tool (`ACTING_TOOLS`, plus any MCP tool whose name sends, creates, updates, deletes or shares) in `gated_handlers` / `gated_call_tool`. The principal-only tools (`UNATTENDED_WITHHELD_TOOLS`) stay withheld, and every guard inside a handler still applies, including department approval levels (a booking in a department on Proposes is still a proposal) and the contact-egress rule. New mail or chat (any turn with an `origin_channel`) wakes the existing reflection early (`take_the_lead.wake`): one run waiting at most, five minutes after the first message and no more than every 30 minutes, as a one-off `take_the_lead_wake` scheduled action. Pause holds it like any scheduled work. **As you** is the Handle it for me dial's top step (delegation section).\n\n**The gate** is plain code (`take_the_lead.check`). Six kinds of action wait for a yes first, each with an Ask first switch, all on to start. The switch is the decision ledger's own mode for the class `take_the_lead_action:<kind>` (`propose` asks first, the default; `auto_execute` doesn't). The kinds are money, contracts and legal, hiring and people decisions, deleting/cancelling/sharing, someone new (an address on neither the roster nor the contacts) and big sends (a broadcast or department message, or more than five people). On top of these, the company (set by the principal) and each person (for their own As you) add structured rules: a person or address, a domain, words, or an amount at or above a figure. Added rules always hold, and an error reading them holds the action. The one thing the gate lets straight through is a direct message (`message_person` or a Slack, Telegram or Discord DM) whose only recipient is the principal (`_only_to_principal`): they are who it would wait for, so a card would only ask them to approve a message to themselves. Anything that reaches or acts on anyone else is checked, and a roster that can't be read is not the principal. A held action becomes a `take_the_lead_action` decision instance with a companion `DECISION_ALERT_SOURCE` card on Today, routed to the first teammate holding one of its People approval ranges (`take_the_lead.approval_ranges`, plain code, closest fit first: money to `spend_lt_2k`, `spend_lt_10k` or `spend_gt_10k` by its largest amount, with a higher range able to approve a lower one and no amount read as large, or to `customer_credit` for credit and payment terms; vendor contracts to `vendor_onboarding` then `legal_sign`, other contracts to `legal_sign`; hiring to `hiring_signoff`; anything else about the board or investors to `board_comms`; in Personal as in team mode), or else the principal through the authority gate's fallback (`departments.authority._route_proposal`). Deletes and shares always go to the principal. Only the principal and that approver see or resolve it (`DecisionClassSpec.principal_and_approver`). Approving it claims the row and carries out the exact call once (`take_the_lead.carry_out`: the unwrapped handler, or the gateway's `call_tool`, on an unattended session, with contact egress only when the principal approves). Declining drops it. Everything done or held is a row in `take_the_lead_log`. What it did and what was approved show in Recent activity (`today._build_activity`, kind `took_the_lead`), named but never quoted, since that feed is the team's. The held card's own `decision_resolved` row is left out because its payload quotes the message. `GET /today/activity` also adds the caller's own Handle it sends (kind `sent_as_you`, actor `you`), which are never in the shared feed.\n\n**In training.** The As the Executive switch has three positions: Off, **In training** and On (`take_the_lead.training`). In training, an action the gate would let through still waits, as a `training` hold for the principal alone, unless the principal has allowed that action with that person or thing (`take_the_lead.allowance`: \"Message Priya Nair\" covers a message to her on any channel; \"Book meetings with Priya Nair and Sam Lee\" meetings with exactly them; a workflow by id; a connected tool by name and the file it names, never one that reaches people such as a mail send or a share). The card carries Approve, **Approve + allow**, **Edit** (the message, a meeting's title, times or description, a task and its due date: `take_the_lead.apply_edits`) and Decline. Allowing is the principal's alone and needs a request the API can tie to them, as turning it on does; the six kinds and the added rules hold whatever is allowed. An edit approved with \"Do it like this next time\" is kept as an example (only the edited text, the principal's own words), and `take_the_lead.learned_note` hands the examples and the fact it's in training to the reflection and research passes in their user turn. After `SUGGEST_AFTER` (3) unchanged approvals of the same action in 30 days, Settings suggests allowing it. What it's learned is one list keyed by feature (`take_the_lead_allowed.feature`), so other features that learn from approvals can join it; each keeps its own switch. Act as me is the second (`act_as_me`, per person, `delegation.training`), with a training switch per setting (Replies, Follow-ups, Suggested actions, Drafts): its rows carry the person's id and the setting in their key, are listed and removed by that person alone, count toward their own cap, and never reach `learned_note`.\n\n**AI disclosure.** Nothing the Executive sends to people outside the web chat goes without saying it is an AI: plain code adds the line in the MCP gateway (mail) and at each chat send (Slack, Telegram, Discord, Google Chat replies, its DMs and room posts), never the prompt, so the persona and the cached system blocks are unchanged and a reply the model writes can't drop it. Replies the inbox watcher writes as a person carry their own line by that person's switches. See Integrations → **AI disclosure** and Act as me → **AI line on replies**.\n",
"mermaid": "flowchart TD\n User([User message]) --> Exec[\"Executive<br/>orchestrator\"]\n Exec -->|consult_specialist| Cso[\"cso: StrategyAgent\"]\n Exec -->|consult_specialist| Cfo[\"cfo: FinanceAgent\"]\n Exec -->|consult_specialist| Gc[\"gc: LegalAgent\"]\n Exec -->|consult_specialist| More[\"chro / coo / cmo / cpo / sales<br/>board_comms / triage\"]\n Mode{{\"Team or solo<br/>(persona + tools per turn)\"}} --> Exec\n Exec -->|\"ghostwrite_email (Act as me)\"| Gw[\"ghostwriter<br/>(tool-less, not a Council agent)\"]\n Cso --> Claude[\"Claude (Anthropic)\"]\n Cfo --> Claude\n Gc --> Claude\n More --> Claude\n Claude --> Exec\n class User entry\n class Exec compute\n class Mode compute\n class Cso compute\n class Cfo compute\n class Gc compute\n class More compute\n class Claude external\n class Gw compute",
- "generated_at": "2026-10-12T13:58:17+00:00"
+ "generated_at": "2026-10-13T15:30:00Z"
}
diff --git a/packages/core/openexecutive/architecture/prebuilt/api.json b/packages/core/openexecutive/architecture/prebuilt/api.json
index 1c8e433b..e264a6ae 100644
--- a/packages/core/openexecutive/architecture/prebuilt/api.json
+++ b/packages/core/openexecutive/architecture/prebuilt/api.json
@@ -1,7 +1,7 @@
{
"section_id": "api",
"title": "API Reference",
- "markdown": "The FastAPI surface lives in `openexecutive.api.main`, which mounts one router per topic. A shared-secret middleware gates every request via an `x-api-key` header, except `/health` and the `/webhook/*` callbacks (which carry their own verification). With local login (`OE_LOCAL_LOGIN=1`, which only `make dev` sets, and only while neither Google nor SSO sign-in is set up) a second middleware answers 403 to any request whose raw `Host` is not `localhost`, `127.0.0.1` or `[::1]` (except a webhook that verifies its caller: Google Chat and Slack always, Telegram only with a `TELEGRAM_WEBHOOK_SECRET` Telegram can send), and to any write whose `Sec-Fetch-Site` is present and not `same-origin`/`none`. The UI signs that owner in with no password and this API usually has no secret, so without it a web page in the owner's browser could drive the API as the principal — by DNS-rebinding its own hostname to 127.0.0.1, or by posting a plain form from another site or localhost port. CORS is restricted to configured UI origins. The gate is off when `BACKEND_SHARED_SECRET` is unset — intended for local development only — so any internet-reachable deployment sets `OE_PUBLIC_DEPLOYMENT`, and `create_app()` then refuses to boot without the secret rather than serving an unauthenticated API.\n\n**Who is calling** is read in one place, `api/caller.py`, which answers with a `Caller` kind and email; a test fails on any other read of an `x-caller-*` header. By default (`open`) it is the `x-caller-email` the UI proxy stamps from the sign-in, trusted because whoever holds the shared secret is trusted as the proxy, and a request with none runs as the principal. **Signed callers** remove that trust. With `CALLER_ASSERTION_PUBLIC_KEYS` set, the proxy signs every request with an Ed25519 key only the UI holds, and a third middleware checks the `x-caller-assertion` header:\n\n- **What it covers.** It is good for one method, the exact path and query, at most 120 seconds, and one use. The API keeps used ids in memory.\n- **What it names.** A `user`, by email, or the `operator`: local login, which runs as the principal.\n- **No caller header at all** is a `service`, for scripts or MCP clients holding only the secret. It is never the principal, not even on an install with no owner yet, so it can't save a first setup.\n- **Anything else is a 401.** That includes an `x-caller-email` without an assertion (`caller_assertion_required`), and a bad, expired, reused or mismatched assertion (`caller_assertion_invalid`).\n\nPublic keys that can't be read stop the boot, rather than fall back to trusting the header. `scripts/make-caller-keys.py` makes the pair, and `scripts/mint-caller-assertion.py` signs one request for curl. The API holds only public keys. Setup's API-protection check warns while a shared secret is set without them.\n\nEndpoints grouped by router:\n\n| Router | Key endpoints | Purpose |\n|---|---|---|\n| `chat` | `POST /chat`, `POST /chat/stop`, `POST /chat/add`, `POST /chat/background` | SSE stream of a live Executive turn. `POST /chat/background` starts a new chat that the server answers to the end with nobody connected (a phone approving a card, then closing the app): it answers `202` with the new `session_id` once the same caller, access and quota checks as `/chat` pass, the turn saves itself as a `/chat` turn does, and one caller may have three running at once (`429` beyond that). `POST /chat/add` hands the running turn a message the person typed while it works; the turn takes it at its next step and the stream names it in a `message_added` event, and a second `POST /chat` on the same conversation waits for the first to finish. The request body optionally carries `page_context` (route, page title, guide section id, and a form descriptor with current values) — sent only by the Ask OE side panel so the Executive can explain the current page or propose form values. When the Executive calls its `propose_form_values` tool, the stream emits a `form_patch` event (`{type, form_id, fields, rationale, iteration}`) that the panel applies into the on-screen form as highlighted suggestions; the user reviews and saves through the form's normal submit path. Each round of tool calls also emits an `activity` event (`{type, label, tool, iteration}`) naming what is running — progress only, forwarded verbatim and never persisted with the message, unlike an `action_taken` chip. A `session_id` in the body must be a chat the caller may use (the same ownership rule as `sessions`). If it isn't, the turn runs in a fresh chat instead, so nothing is written into someone else's session and the reply looks the same as for an unknown id. An unknown namespaced id (`slack:…`, `telegram:…` and so on) is ignored and a fresh chat is started, because those ids belong to the channel adapters and claiming one first would make the caller its owner. The client mints a `client_turn_id` and sends it with the turn; `POST /chat/stop` takes that id and halts the turn mid-flight, so the button is live from the moment Send is pressed rather than only once the first SSE byte arrives. A stopped turn emits a `stopped` event and then the usual `done` over the same still-open stream — never an `error` — and whatever streamed so far is persisted and flagged. The endpoint returns 404 both for an unknown id and for another caller's turn, so it cannot be used to probe which turns are live. An optional `memory_text` (1–2000 chars) is what peer memory records as the caller's words instead of `message` — the briefing handoffs send it because their seed quotes the Executive's own card; `POST /chat/upload` derives it from the typed text and filenames. Since it never reaches the transcript, the `chat_turn` audit row keeps it beside `message`. After the reply — a stopped or timed-out one too, before the `stopped` / `error` / `done` frames — a turn that looked at anything sends one `sources` event, `{type, session_id, sources: [{kind, title, url}], unavailable: [area]}`: `kind` is `company`, `knowledge`, `notion`, `research`, `document` or `web`, `url` is only ever an http(s) link or, for an earlier document, `/artifacts/<id>`, at most 12 sources (6 from the web, pages the reply cites first), and `unavailable` names parts of the analysis (\"finance\") whose specialist failed or said nothing every time it was asked. It is persisted with the reply as `chat_messages.sources`. A request that carries `x-caller-email` marks its turn signed in (`Session.web_caller_signed_in`), which Act as me needs on a deployment without local login. |\n| `sessions` | `GET /sessions`, `GET /sessions/{id}`, `GET /sessions/{id}/messages`, `DELETE /sessions/{id}`, `POST /sessions/{id}/messages/{message_id}/feedback`, `GET /sessions/{id}/followup` | List, read and delete saved chat sessions (assistant messages carry their row `id` and any `feedback`). The list is scoped to the caller, and each chat carries `preview`: its last message as one plain line, Markdown marks stripped and cut at a word near 120 characters, with `You:` before the person's own words. With `?q=` it is a search: the caller's own chats whose title or message text contains the words (case-insensitive, `LIKE` with wildcards escaped), each with `match_count` (messages that contain them; 0 for a title-only match) and `snippet` (the newest matching message, around the match, prefixed `You:` or `Executive:`). Scoped exactly like the list, so nobody else's conversation can match. The Chats page searches this way and opens a result with `?find=`, so the chat starts on that message and steps through the others. A chat-app conversation started with /new is left off the list until its first turn (`memory.conversations`). Every per-session route checks ownership, because session ids are guessable (`slack:dm:<user id>`, `telegram:<chat id>`): only the session's owner or the principal gets through, and a legacy row with no owner is the principal's alone. A teammate's own chat (a web chat, a DM, a private Telegram chat; not a channel or group conversation, `chat._shared_space`) is theirs alone, the principal included, as their mail with the Executive is. Anyone else gets the same 404 as for an unknown id, so the routes can't be used to find out which chats exist. The one other way in is having started the chat in this process (matched by Person id or verified email, so it survives the caller being added to the roster), which is how a caller the roster can't resolve keeps using their own chat. That record is in memory, so after a restart such an ownerless chat is refused for reads. The principal can continue someone else's ownerless chat without becoming its owner. Deleting a session also drops its in-memory copy, so a later turn naming that id starts fresh. The feedback route records 👍/👎 (`{\"feedback\": \"up\"|\"down\"|null}`) on one assistant reply; only the session's own caller or the principal may rate it. The rater is stored (`feedback_by_person_id`), and a 👎 on a reply to the rater's own message triggers a working-style pass for them (paced; see attunement). The chat stream's `done` event carries `message_id` for the persisted reply. The followup route returns `{\"suggestion\": string|null}` — one suggested next user message, written by the utility-fast model from the last few turns, that the chat composer shows as its greyed placeholder (Tab / → or the inline chip fills it in without sending). It is null unless the session ends on a persisted assistant reply, and null (uncached) when the model fails. Same ownership check as feedback; successful suggestions are cached in-process per (session, reply id). Messages from `GET /sessions/{id}/messages` carry each assistant reply's saved `sources` (the same shape as the chat stream's `sources` event, without `type`), so a reloaded chat shows them again. |\n| `today` | `GET /today`, `GET /today/activity`, `GET /today/activity/daily`, `GET /today/brief-delivery`, `GET /today/top-three`, `GET /today/weekly-review` | Morning-brief aggregation, recent-action rail, and Pulse heartbeat heatmap (per-day activity counts). `GET /today` carries the viewer's cached \"What's going on\" `narrative` with `narrative_generated_at` (when it was written) and `narrative_stale` (a fresher one is being written — the page re-polls; see Today). A proposal backed by a roster request carries `roster_request` (the new sender's server-derived details and held-message previews; the principal's only) and is answered through the `people` routes below. `brief-delivery` returns the latest morning brief or end-of-day digest that didn't reach the owner, as `{brief, at, problem, fix, readable}` in plain language (`readable` is false when it couldn't be written, so there is nothing to read), or null — null while briefs go out, once the cause is fixed, while the scheduler is off, and for anyone but the owner (`chat._caller_is_principal_or_unclaimed`); it carries no address or channel id. `top-three` and `weekly-review` feed the solo Briefing's own cards and answer only in solo mode and only to the principal (`workflows._caller_is_the_principal`: a principal match or no caller header, never an unclaimed roster); anyone else, and team mode, get null. `top-three` returns `{items: [{key, kind, text, why, slot}]}` — the morning brief's top three from the same code (`build_top_three(principal_due_soon())`), `slot` a free block like `10:00–11:00`, `\"\"` when none is left, null when no calendar was read — and is kept off `GET /today` because its calendar read may take up to 4 s. `weekly-review` returns the latest completed `weekly_review` run as `{run_id, completed_at, period, top_three, excerpt}` (next week's top three as plain text; `excerpt` when that list is empty), or null when none has completed. |\n| `people` | `GET\\|POST\\|PATCH /people`, `POST /people/{id}/archive`, `GET /people/requests`, `GET /people/requests/{id}`, `POST /people/requests/{id}/approve`, `POST /people/requests/{id}/decline`, `GET\\|POST /people/{id}/open-loops`, `POST /open-loops/{loop_id}/close`, `GET /people/{id}/outreach`, `GET\\|PUT\\|DELETE /people/{id}/attunement` | Roster CRUD and scope lookups. Every `Person` carries `kind` (`team` / `contact`). `GET /people` lists team members only unless `?include_contacts=true` (the People page asks for both; the department-head and workflow pickers keep the default); `POST` / `PATCH` accept `kind`, and a principal contact is a 422 either way. They also take `email_aliases` (other addresses the person writes from, at most 10; on `PATCH` the full list replaces the current one, omitted leaves it): an alias matches their mail and may be emailed but never signs in; a malformed one is a 422, and an alias — or a new primary `email` — that is already someone else's address or alias is a 409. Every `POST` / `PATCH` then closes any pending roster request whose sender now matches someone and replays its held messages. Adding, editing and archiving (`POST`, `PATCH`, `/archive`) are the principal's only, or anyone's while no one is principal yet so a first setup can add its owner (`chat._caller_is_principal_or_unclaimed`, the check behind `PUT /workspace`; a request with no `x-caller-email`, such as the CLI or local login, counts as the principal). Anyone else, a person editing their own row included, gets 403 before the person lookup, so it can't probe which ids exist. Contacts are private to the principal (`caller_is_principal`: the signed-in email is the principal's, or no `x-caller-email`; stricter than the owner rule above, so while no one is principal nobody is): for anyone else `include_contacts` is ignored, a contact's id is a 404 on `GET /people/{id}` and open-loops (exactly as for a missing id; `PATCH` and archive are already a 403 for them), and creating a contact or turning someone into one is a 403, even on a first setup with no principal yet. `GET /people/me` returns `{person_id, is_principal}` so the UI offers contacts to the principal only. `GET /people/by-scope/{token}` never returns a contact. Archiving a person closes their open loops. The open-loop routes list what a person owes (soonest due first) and close one (`reason`: done / not_needed / cancelled) — both the principal's or the owner's only (403 otherwise, via `people.store.is_principal_or_self`). `POST /people/{id}/open-loops` `{task, due_date?}` assigns that person a task (201 with the loop; `due_date` is a local date, due 17:00 in the user's timezone, clamped to 60 days, default 2 days): any active team member may assign anyone on the team, 403 for anyone else; 409 for an archived person, a contact (as seen by the principal), a person at the open-loop cap, a duplicate or open loops turned off; 422 for an empty task. Nobody is messaged at assignment — the owner is chased once it is due. `/outreach` returns, per kind of proactive outreach over 30 days, `sent` / `replied` / `acted` / `ignored` / `pending` (voided DMs are excluded) — same access rule, checked before the person lookup so it can't probe which ids exist. `/attunement` reads, replaces (`rules`: up to 4 strings, or omitted to only set `locked`) or resets the person's working-style rules — same access rule, 404 for an archived person; `PUT` rejects a rule that is not purely about style with 422. Archiving a person also drops their working style and frees their aliases. The **roster-request** routes (declared before `/people/{person_id}`) are the principal's alone (`caller_is_principal`; anyone else gets 404, as for a missing id): `GET /people/requests?status=pending|approved|linked|declined|expired|superseded|all` lists up to 100 (`id`, `channel`, `channel_ref`, sanitised `display_name`, `profile_email`, `on_company_domain`, `suggested_kind`, `suggested_person_id` / `_name`, `status`, `resolved_person_id` / `_kind`, `message_count`, timestamps, `ack_sent`, and `previews` of the held messages while pending — shown to the principal, never to a model); `approve` takes either `link_person_id` (+ `replace_channel_id`) or `full_name` + `kind` (no default: 422 without both) + optional `role`; `decline` takes no body. An answered request is a 409, a chat id or address already on someone else is a 409; the held messages are replayed (or dropped) in the background. |\n| `departments` | `GET\\|POST\\|PATCH\\|DELETE /departments`, `.../goals` | Department state and per-department goals. `PATCH /departments/{slug}` also takes `watched_entities` (≤50 names, `[]` clears) — the entities the research watch policy may add watches for and route to this department. A `cadences` spec that does not parse is rejected with a 422 (empty means no cadence). A new `head_person_id` must be a team member: an unknown id and one of the principal's contacts get the same name-free 422 (`person_id N is not on the team.`), so the route — open to any signed-in user — never reveals that a contact exists; an unchanged head is not re-checked. `POST .../goals` requires only `key_result`: a missing or blank `period_value` is filled with the current period for `period_type` in the user's zone (the same label `create_goal` uses), and `target` defaults to empty. `PATCH .../goals/{id}` rejects a blank `key_result` or `period_value` with a 422; `target` may be cleared. |\n| `agents` | `GET /agents`, `GET /agents/models`, `GET /agents/models/options`, `GET /agents/presets`, `POST /agents/presets/{id}`, `PATCH /agents/{id}`, `POST /agents/{id}/test` | Specialist metadata and Council overrides. `/agents/models` is the flat model allowlist the PATCH validator uses; `/agents/models/options` returns the same ids in the same order, each tagged with `provider`, `provider_label`, `route` (`direct` / `openrouter` / `local`) and a display `label`, for the Council's Provider → Model picker. `AgentDetail` carries `instructions` (additional instructions, appended after `prompt`, which stays the built-in prompt or its replacement); `PATCH` and `/test` accept `instructions` (max 4,000 chars, blank clears it). `GET /agents/presets` lists the Fast, Balanced and Thorough quality presets (each with `available` and the resolved `model` for this install) plus `active` (null when the council is Custom), `base` and `custom_agents`; `POST /agents/presets/{id}` applies one by writing per-agent model and deep-reasoning overrides and returns the same shape (404 unknown preset, 409 when the install allows none of its models). Each `AgentMeta` carries `visibility` (`core` or `internal`) so the Council's simple view can list core agents first. |\n| `personas` | `GET\\|PUT\\|POST\\|DELETE /personas` | Executive persona definitions. |\n| `alerts` | `POST /alerts/{id}/ack`, `POST /alerts/bulk-ack`, `POST /alerts/{id}/reopen`, `POST /alerts/review` | Groom the briefing queue. `ack` (`read`/`ack`/`dismissed`, optional `mute_topic`) also teaches a watch-sourced alert's watch (`trust_score`); `bulk-ack` takes explicit `alert_ids` and/or `older_than_days` + `category` (exempt sources skipped); `reopen` is the Undo for a `resolved` / `expired` / `dismissed` alert (never `ack` or an artifact / decision row; back to `unread` with every review stamp cleared and its TTL anchor refreshed, so it is live again even if older than its TTL); `review` runs the Executive's relevance review on demand (ignores the interval; still honours `ALERT_REVIEW_ENABLED`, the per-pass caps, the single-flight lock and per-alert idempotency) and returns the move counts. `bulk-ack` caps ids at 500, needs `older_than_days >= 1`, and applies watch feedback per row; `mute_topic` must be one of the alert's tags or a scoped `external:` / `department:` pattern. Alerts still surface only through the briefing (`/today` reads the live view) — there is no list, prefs, mutes, or live-stream endpoint. |\n| `artifacts` | `GET /artifacts`, `GET /artifacts/{id}`, `GET /artifacts/{id}/download?as=`, `POST /artifacts/{id}/archive`, `POST /artifacts/{id}/restore`, `DELETE /artifacts/{id}` | Browsable deliverables in any format — Markdown, HTML, Word, Excel, or a link into a connected app — with `format`, `downloads`, `external_url`, `supersedes_id` and, for an extension's kind of document, `collection` / `collection_label` on every summary (`orchestrator.extensions`; archive, restore and delete tell the collection first and answer 502 if it refuses). `download` renders the file (Word/Excel built at request time from the stored text; any Markdown artifact, workflow output included, can export `?as=docx`) and is always an attachment with `nosniff` and a sandbox CSP. Every route answers for the signed-in caller (`artifact_records.Viewer`): a drafted artifact is its owner's alone (`alerts.owner_person_id`, the person whose conversation published it; none = the principal's), a workflow output follows its run (`persistence.run_visible_to`), and anything the caller may not see is a 404 like a missing id, the principal included. `POST /alerts/{id}/ack` and `/reopen` answer 404 for someone else's drafted artifact the same way. |\n| `decisions` | `GET /decisions`, `POST /decisions/{id}/{approve,reject,cancel}`, `GET`/`PUT /decisions/classes/meeting_scheduling`, `GET /audit/reliability` | Approval-gated decision instances (e.g. calendar bookings). What approve/reject/cancel do is per class (`DECISION_CLASSES`, one `DecisionClassSpec` each; meeting booking books on approve and deletes on cancel; `delegation_reply`, an inbox reply card, is its approver's alone (`approver_only`) — hidden from everyone else, the principal included when it is a team member's — sends its Gmail draft on approve and, after a reject, deletes the draft unless it was edited), and an instance of a class with no spec gets 409 instead of being carried out as another. Only the principal, or the person the decision went to for approval (`approver_person_id`), may approve, reject or cancel it. Anyone else gets 403, and a decision nobody was asked about, or one private to the principal, is the principal's alone. The resolver is recorded as `resolver_person_id`, and a reject that loses the race to another answer is a 409. `GET /audit/reliability` is the principal's alone (403 otherwise), because it counts private decisions too. A proposed booking is surfaced on the briefing as a companion alert; approve/reject/cancel here also clear that alert. An approval whose caller resolves to the principal runs the booking inside `grant_contact_egress()`, so a meeting the principal proposed with a contact can be created. Such a booking's payload is marked `private`: it is proposed to the principal, left out of anyone else's `GET /decisions`, and a 404 for them on `GET /decisions/{id}` and every action route. Reviewed in the briefing — there is no separate proposals page. The class routes read and set how meetings are handled: both return `{decision_class: \"meeting_scheduling\", mode: \"propose\" \\| \"auto_execute\"}` (default `propose`); the `PUT` body is `{mode}`, principal-only (403 otherwise, open before a principal exists — but then only `propose` can be set: `auto_execute` is a 409 until a non-archived principal is on the roster), 422 on any other mode, and a change writes a `decision_class_mode_changed` audit row. In solo mode the class mode decides a booking, and `auto_execute` applies only to a turn the principal started on a verified surface — any other turn gets a proposal. |\n| `delegation` | `GET`/`PUT /delegation`, `GET`/`PUT`/`DELETE /delegation/voice`, `POST /delegation/voice/learn`, `POST /delegation/voice/describe`, `POST /delegation/voice/signature`, `PUT /delegation/inbox`, `POST /delegation/inbox/check`, `GET /delegation/replies`, `PUT /delegation/replies/{id}/text`, `GET /delegation/actions`, `PUT /delegation/handle-it`, `PUT /delegation/training`, `DELETE /delegation/learned/{id}`, `GET /delegation/handled`, `PUT /delegation/team` | Act as me — the **caller's own** delegation only; nobody sets it for anyone else. It answers the principal, and a team member once the owner lets them; anyone else gets 403 `not_available_yet`. `GET /delegation` returns `{enabled, gmail: {status, message, email, connect_command, outlook_connect_command, provider}, inbox: {enabled, status, message, watch_since, last_poll_at, checking}, handle_it: {enabled, mode, available, sent_today, lead, lead_available}, training: {replies, follow_ups, actions, drafts, learned}, team}`, where `team` is the owner's switch for the principal while `DELEGATION_TEAM_MEMBERS` allows it — `{enabled, members: [{person_id, name, enabled, inbox, drafts_30d, sent_30d}]}`, counts only — and `null` otherwise. `PUT /delegation/handle-it` takes `{enabled?, mode?}` (`mode` is `careful`, `balanced` or `bold`; anything else is a 422 `bad_mode`) from a caller the API knows is that person — 403 `not_yours`, 409 `caller_signing_required` with only the shared secret — and turning it on needs Act as me and Draft replies to my inbox on (409 `act_as_me_off` / `inbox_off`); `GET /delegation/handled` returns `{replies: [{decision_id, sent_at, to_name, to_email, subject, body, open_questions, gmail_link, source}]}`, the caller's replies and follow-ups (`source` `follow_up`) sent on its own in the last 7 days; a card on `GET /delegation/replies` carries `source` the same way, `can_allow` while Replies (Follow-ups, on a follow-up) is in training, and `learns_style` while Drafts is. `PUT /delegation/training` takes `{replies?, follow_ups?, actions?, drafts?}` (any other field is a 422) and returns the `GET /delegation` shape; putting Replies or Follow-ups in, or Suggested actions out, needs no signed sign-in, anything else does (409 `caller_signing_required`), and putting Replies or Follow-ups in turns Take the lead as you off (`PUT /delegation/take-the-lead` on answers 409 `in_training` while they are). `training.learned` is `[{id, label, setting, example, uses, created_at}]`, what the caller allowed or kept in training (`setting` `replies`, `follow_ups`, `actions` or `drafts`; the kept example, or \"\"); `DELETE /delegation/learned/{id}` removes one of the caller's own (404 for anyone else's). `PUT /delegation/team` takes `{enabled}` from the principal alone (403 `principal_only`; 409 `not_available` when the install doesn't allow it). The Gmail `status` is `connected`, `not_configured`, `needs_reconnect`, `mismatch`, `no_email`, `shared_mailbox` or `error`, checked live with Google on each call. `PUT` takes `{enabled}` (any other field is a 422); turning it on needs the caller's own Gmail connected, otherwise a 409 whose code says why (`gmail_not_connected`, `gmail_needs_reconnect`, `gmail_mismatch`, `no_email`, `shared_mailbox`, `gmail_error`). The voice routes show, learn (from the caller's sent mail, one pass at a time — a 409 `in_progress`, `locked`, `too_soon`, `not_enough_mail`, `no_profile` or `changed` when it was locked or edited while learning, or a Gmail code), write it from the caller's own description (`POST /delegation/voice/describe` with `{description}`: one model call applies their words to the profile they have and returns it with a short `sample_reply`, saving nothing, so they review it and save it with `PUT`; a 409 `empty`, `too_long`, `in_progress`, `too_many` past 10 an hour or `no_profile`; no mailbox needed), edit and lock, and reset \"How I write\": a typed field that fails validation is a 422 `invalid_voice` listing each `rejected` field, and `clear_exemplars` / `clear_signature` drop those. `POST /delegation/voice/signature` reads the signature from the caller's Gmail settings again and changes nothing else (a locked profile stays locked), with the same Gmail codes as learning: a 409 when their mailbox isn't connected, it refuses the sign-in or the check itself fails (`gmail_error`), and a 502 `gmail_error` when reading the signature fails. On an Outlook mailbox it is a 409 `signature_unavailable`, since Outlook gives apps no signature to read, and the stored one is left alone. A caller `can_delegate` refuses — anyone but the principal, for now — gets 403 `not_available_yet`, which hides the Settings card. A request with no `x-caller-email` is a 403 `sign_in_required` unless the API runs under local login — a header-less call is not a sign-in (the caller is read through `api/caller.py`, as on every principal-only route). `PUT /delegation/inbox` takes `{enabled}` and turns **Draft replies to my inbox** on or off, returning the same shape as `GET /delegation`: on needs Act as me on (409 `act_as_me_off`) and the mailbox connected (the Gmail codes above, which keep their names for an Outlook mailbox too), and turning Act as me off turns it off too. `POST /delegation/inbox/check` starts a scan now in the background and answers 202 with the `inbox` state (409 `inbox_off` or `in_progress`). `GET /delegation/replies` returns `{cards: [...]}`, the caller's open reply cards from the database alone (no Gmail call): the decision id, the sender's name, address, relation and whether it was verified, the subject, when it arrived, what they wrote, the draft's recipients, subject and body, `open_questions`, `flags` and a Gmail link built from a fixed prefix. **Send** is `POST /decisions/{id}/approve`: it sends the card's draft exactly as it is in Gmail, after checking again that the caller is its owner — which on a server needs signed callers (409 `caller_signing_required` otherwise) — that both switches are on (`inbox_off`), no client slot is active (`client_slot`), the Gmail is connected (the Gmail codes), the draft is still there and unanswered (`draft_gone`, `you_replied`) and its recipients are allowed (`no_recipients`, `too_many_recipients`, `executive_recipient`, `not_from_you`). Recipients that changed, or a newer message in the thread, answer 409 `{code: \"confirm\", message, reasons, recipients}`; approving again with `edits: {recipients: <the list shown>, thread_moved_on: true}` is the second yes. **Send + allow** adds `edits.allow: true`: 422 `cant_allow` unless Replies (Follow-ups, on a follow-up card) is in training, 409 `learned_full` when their list is full, both before anything is sent; `edits.example: true` keeps a draft edited in the mailbox as how they write to that person while Drafts is in training. A draft edited in Gmail between those checks and the send is not sent (`draft_changed`). A double tap sends once (`already_handled`); Gmail refusing answers 429 `rate_limited` or 502 `gmail_error` with nothing sent, and a send Gmail didn't confirm answers 502 `send_unconfirmed` and is settled by the next inbox scan. Dismiss is `POST /decisions/{id}/reject`. `GET /delegation/actions` returns `{cards: [...]}`, the caller's own approval cards (`delegation_actions`): the decision id, when it was made, a one-line `why`, and each action's `index`, `kind` (`invite`, `message`, `add_contact`), `summary` and `text` (a message as it will be sent, or a meeting's description). **Approve** is `POST /decisions/{id}/approve` with `edits: {only: [indexes]}` (and `allow: true` for Approve + allow while Suggested actions is in training: 422 `cant_allow` otherwise or when nothing on it can be allowed, 409 `learned_full`; a card carries `can_allow`); it needs the card's own person signed in (403 `not_yours`, 409 `caller_signing_required`), Act as me on (409 `act_as_me_off`) and a card under a week old (409 `expired`), runs once (409 `already_handled`), and returns the decision with `final_payload_json.results`: per action `done`, `waiting` (sent to whoever approves meetings), `failed` with why, or `skipped`. Every change writes a private audit row. |\n| `history` | `GET /memories/history` (`?q=`), `PUT /memories/history/settings`, `PATCH`/`DELETE /memories/history/{id}`, `POST /memories/history/{id}/forget-conversation` | Always in the loop: the **caller's own** notes and switches only, resolved as for Act as me (a signed-in person on the roster, or the principal under local login; otherwise 403 `sign_in_required` / `not_on_roster`). `GET` returns `{notes, reply_notes, retention_days, effective_retention_days, company_retention_days, retention_choices, can_keep_notes, can_note_replies, can_set_company_retention, share_work_style, can_share_work_style}` (`can_keep_notes`: a team member on the People list may turn notes on; `can_note_replies`: they can use Act as me, so email replies are noted too; `share_work_style`: their own \"Share my work style with the team\" switch, `can_share_work_style`: a team member in a team workspace). `PUT` turns notes on only for someone who may keep them, and `share_work_style` on only for someone who may share it (403 `not_available` for either; turning either off always works), takes a person's retention only when shorter than the company's (422 `invalid_setting`), and the company retention only from the principal (403 `principal_only`). `PATCH` pins or corrects a note; `DELETE` forgets it; `forget-conversation` is \"Don't remember this\". A note that isn't the caller's is 404. Every change is audited privately to its person. |\n| `audit` | `GET /audit/logs`, `GET /audit/logs/{id}`, `GET /audit/sessions/{id}`, `GET /audit/usage`, `POST /audit/log` | Durable audit trail of side effects. `/audit/sessions/{id}` adds a per-session token cost summary; `/audit/usage` aggregates token usage and OpenRouter cost across all sessions (totals, by day, by model, and by source — the part of the system that made the calls — including server-side web-search counts), plus `scripts`: how many `run_script` scripts ran (and worked, and in workflows), the tool calls they made, their time, and `turns_avoided` — an upper bound on the model turns those calls would otherwise have needed (each call past a script's first) — and `python_jobs`: how many Python jobs ran (and worked, ran a kept Custom tool, took a chat attachment), their time, CPU, largest peak memory and bytes in and out, and the tokens and charge of the turns that ran one. The Token usage page shows it as a Scripts section. Rows private to the principal (`private_to_principal`: a turn about their private mail, or a row naming one of their contacts on their own turn) are theirs alone: for any other caller, resolved as for `/alerts`, the usage totals leave them out. Once a principal is on the roster, anyone else reads only their own rows (a teammate's own turns, their Act as me work) in the list, by id and in a session, and the principal reads those as actions only — what was done, when and in whose conversation, with the text, details and `full` left out, and no text search over them; any other row's id is a 404, and a session shows only the rows the caller reads. |\n| `architecture` | `GET /architecture/sections`, `.../sections/{id}` | This documentation surface. |\n| `guide` | `GET /guide/sections`, `.../sections/{id}` | Plain-language user-guide content surface — mirrors `architecture` (static, version-controlled files; no LLM on the serving path). |\n| `health` | `GET /health`, `GET /health/honcho` | Liveness probes. Only `/health` is unauthenticated; `/health/honcho` sits behind the gate like everything else, because the exemption is an exact-path set. `/health` also carries `google_tools_missing`, the pinned Google tools startup priming could not find: `[]` once it has checked and they are all there, a list of names when a workspace-mcp rename or the configured services and tool tier left one out, and `null` for no answer — Google Workspace is not configured, or the probe has not finished or it failed. Only `[]` asserts they are present. It also carries `gateway`, the MCP gateway's state from its supervisor (`orchestrator.gateway_boot`): `state` is `off` (MCP not enabled), `no_servers` (enabled, the config names none), `starting` (an attempt is running), `up` or `down` (every attempt so far failed; `next_retry_at` is the next one), with `attempts`, `up_since` and `email_poller` (whether the poller that rides on the gateway runs); the failure's own text stays out, since the route is open to anyone, and goes to the log and the owner's notice on Home. A hosted instance whose gateway never came up used to answer `ok` here with no email going out or in. Only `[]` asserts the tools are present, so a probe that died cannot read as healthy; Setup status says which of the three `null` means. |\n| `version` | `GET /version` | The running version, the latest release on GitHub and `update_available`, for Settings → About. The GitHub answer is cached 6h, only a semver `tag_name` is used and the release link is built from it; `UPDATE_CHECK_ENABLED=false` skips the call and still returns the running version. No in-app updater: upgrading is host-side (`docs/deployment.md#upgrading`). |\n| `setup_status` | `GET /setup/status` | Settings → Setup status: one light per part of the install — AI model, company setup, owner, the Executive's address, API protection, Gmail, your own mailbox (Act as me), Slack, Discord, Telegram, Google Chat, the scheduler, the daily brief (when and where it goes) and Honcho — as `{checked_at, checks: [{id, label, state, summary, fix, link, last_activity}]}`, where `state` is `ok`, `warn`, `error` or `off` (optional and not set up), `fix` says what to do in plain language, `link` is an in-app page that helps (`/people`, `/onboard`) and `last_activity` is the channel's latest `integration_inbound` audit time. It tests live, with at most one free read-only call per service, each bounded to 5 s: Anthropic `models.list` (so an organisation key missing `ANTHROPIC_WORKSPACE_ID` shows up), Slack `auth.test`, Telegram `getMe` and `getWebhookInfo`, Discord `users/@me` only while the bot is still connecting, and — once the owner's own Gmail credential exists — Google's token refresh and Gmail `users/me/profile`. It also reads the listeners the lifespan leaves on `app.state` (Slack's socket, the Discord bot and its task, the MCP gateway), the scheduler heartbeat and the pause switch. When `google_workspace` is configured, the lifespan also starts `MCPGateway.prime_pinned_tools()` in the background (held on `app.state.mcp_prime_task`, cancelled at shutdown), so the pinned Google tools named in the prompt are callable without a search. With `DRIVE_SYNC_ENABLED` the lifespan also bootstraps the `drive_sync_scan` heartbeat beside Notion's (see Scheduler); the Drive folder sync has no HTTP route, only the `openexecutive sync-drive` / `purge-drive` CLI. With `CONFLUENCE_SYNC_ENABLED` it bootstraps the `confluence_sync_scan` heartbeat the same way; that sync is run by hand with `openexecutive sync-confluence` / `purge-confluence`. Sample values copied from `.env.example` are named as such without being sent anywhere. No token reaches a response or a log line: probes report a status code, a service's short error code or an exception's class name. Requests within 5 s of a run get that run's answer and concurrent requests share one run (kept per app on `app.state`), so the button can't become a stream of calls to those services. Gated like every other route; the UI adds a sign-in light of its own, from its own settings. |\n| `documents` | `POST\\|GET\\|DELETE /documents`, `GET /documents/{filename}`, `GET /documents/sources`, `POST /documents/sources/{drive\\|onedrive\\|notion}/sync`, `GET /documents/{drive\\|onedrive\\|notion}`, `GET /documents/{drive\\|onedrive\\|notion}/{id}` | Upload, list, view extracted text, and delete indexed company docs. `GET /documents` rows carry `source: \"upload\"` and the `domain` the upload was indexed under. The connected-source routes back the knowledge page's single Company documents list: `/documents/sources` gives each source's `{id, label, enabled, syncing, last_run, last_error, interval_minutes, file_count}`; `/documents/drive`, `/documents/onedrive` and `/documents/notion` list synced files from the sync state (`{enabled, last_run, last_error, files: [{id, name, url, modified_at, synced_at, indexed}]}`); `/documents/{drive\\|onedrive\\|notion}/{id}` returns the stored text (a OneDrive id is `<drive id>:<item id>`) for an id on record (the path comes from the state, never the caller). They are read-only — a synced file is removed where it lives. `POST …/sync` starts one tick of the same sync the scheduler runs as a background task (202); 409 when the source is not connected or a tick is already running (every sync holds a per-process run lock, so a manual run never overlaps the heartbeat; the route also keeps its own started task per source and re-checks just before starting one, so two simultaneous requests start one tick, not two), 429 with `Retry-After` within 60 s of the last successful run or of the end of the last tick in this process, failed ones included, so a failing or slow source can't be re-run back-to-back. The page's text view never loads images from synced content (a remote image would be a read beacon for whoever edits the shared folder). Upload is keyed on the filename, so re-uploading a document replaces its chunks rather than duplicating them, and `DELETE` removes the rows as well as the file. `domain` must be one of the nine specialist domains (including `sales`) or `general`; anything else is a 400 rather than a 200 that indexes where no specialist can retrieve. A scanned PDF is converted (`knowledge.pdf_reader`) before it is indexed and in the text view, as are scanned PDFs in `POST /chat/upload` and the client / onboarding intake uploads. The upload is streamed to disk 1 MB at a time and refused with a 413 once it passes 50 MB, and the file is parsed in a short-lived child process (see the RAG section), so a large document does not stay in the API process's memory. A Word or Excel file that cannot be parsed, or whose parse runs out of time, is a 422 with a readable reason. When every parser is busy, the response is a 503 with `Retry-After`. |\n| `executive` | `GET /executive/status`, `POST /executive/pause`, `POST /executive/resume` | Operator pause switch for all autonomous work (scheduler, Gmail poller, workflow resumer); inbound chat and DMs stay live. Status returns `{paused, paused_at, paused_by, reason, held_actions, can_resume}` — `held_actions` counts due `pending` rows that fire on resume. `pause` takes an optional `reason` (≤200 chars) and is idempotent (re-pausing keeps the original start and reason); `paused_by` is the UI proxy's `x-caller-email`, else `api`. Behind the shared-secret middleware — not `SCHEDULED_ADMIN_TOKEN`, which the UI cannot send. Any caller may pause; `resume` is principal-only (`is_principal_or_self`, 403 otherwise; open while no principal is rostered; a header-less caller resolves to the principal as in `chat`), and status carries `can_resume` for the viewer so the UI hides Resume from everyone else. |\n| `workspace` | `GET /workspace`, `PUT /workspace` | Solo/team mode, the user's time zone and the principal's role (`memory.workspace_settings`). Both return `{mode, timezone, effective_timezone, role_kind, role_title, reports_to, remit, measured_on, company_domains, company_domains_custom}` — `timezone` is what the user set (null when unset), `effective_timezone` the zone in effect (`timezone`, else `USER_TIMEZONE`, else UTC), and each role field is null until set. `PUT` is a partial update: only the fields sent change; `mode` is `solo` or `team` (not null), `timezone` an IANA zone, or null/blank to clear it; `role_kind` is `owner`, `in_house`, `independent` or `other`, and `role_title` / `reports_to` / `remit` / `measured_on` are free text (trimmed, at most 120 / 120 / 500 / 300 characters), each null or blank to clear it; an unknown zone, a bad mode or role kind, an over-long role field or an unknown field is a 422 and changes nothing. Switching to solo cancels every pending department check-in, switching to team re-schedules them, and a new zone re-times the pending briefs and reflection; a role change has no side effects (the next turn reads it). `company_domains` is the company's own email domains (a teammate matches by local part on them — see Org): `GET` returns the ones in effect (the stored list, or the domain of the principal's primary address, with `company_domains_custom` false) only to the same caller the role is shown to, `[]` for anyone else; `PUT` takes a list (at most 10 plain domain names, never a free-mail provider's — 422 otherwise, the value never quoted) or null / `[]` to go back to deriving them, and a change is audited with the new list. Each change writes a `workspace_settings_changed` audit row (a role change adds `details.role_fields_changed`, the names of the fields that changed — never their text). The role is the principal's: `GET` returns it only to a caller `PUT` would accept and reads it as nulls for anyone else. `PUT` is the principal's alone (`is_principal_or_self`, 403 otherwise; a header-less caller resolves to the principal as in `chat`) and open while no principal is rostered, so first-run setup can pick a mode. Both work before onboarding — unlike `company-profile`, nothing here needs a profile. |\n| `scheduled` | `GET\\|DELETE` scheduled actions (`?status=…&order=asc\\|desc`) | Pending and fired scheduled actions; `order` flips the `run_at` sort (asc=soonest-first for the upcoming queue, desc=most-recent-first for terminal-status history). `DELETE /scheduled/{id}` cancels a `pending` row (409 otherwise). Like pause it only stops work, so any signed-in user may — the UI proxy's `x-caller-email` is trusted when the request also carries the `BACKEND_SHARED_SECRET` `x-api-key` — as may a matching `X-Admin-Token` (`SCHEDULED_ADMIN_TOKEN`) or a loopback caller with no token set; otherwise 401 (token set) or 503. Audited as `scheduled_action_cancelled`. |\n| `watchlist` | `GET\\|POST\\|PATCH\\|DELETE /watchlist`, `POST /watchlist/{slug}/approve`, `POST /watchlist/{slug}/decline` | External-monitoring watchlist + signals. Rows carry `origin` (`manual` \\| `executive` \\| `research` \\| `research_proposed`); `approve` turns a research suggestion (`research_proposed`, `dry_run`) into a live watch, `decline` `{reason: not_relevant \\| too_noisy \\| wrong_source}` removes it and records the decline (or, for `too_noisy`, keeps it live at a high floor), both 409 unless the row is a pending suggestion; `DELETE ?reason=` on a research-origin row records the decline too. |\n| `workflows` | `GET /workflows`, `POST /workflows/{name}/runs`, `GET\\|POST\\|PUT\\|DELETE /workflows/custom`, `GET\\|DELETE /workflows/custom/{name}/targets`, `POST /workflows/runs/{run_id}/decision`, `GET /workflows/tools/search?q=`, `GET /workflows/tools/describe?names=` | Workflow catalog, run management, and user-created (dynamic) workflow CRUD. Custom definitions come back with a server-managed `owner_person_id` (the creator — the web caller on `POST`, set on insert only; a client-sent value is ignored), who is asked to approve a workflow's first write to a new target. `GET /workflows/custom/{name}/targets` lists the targets its tool steps may write to without asking and `DELETE …/targets?value=` forgets one. `POST /workflows/{name}/runs` is a 403 (\"Only the principal can run this workflow.\"), audited, before any run row exists, when the workflow is one only the principal may start in the workspace's mode (`weekly_review` always, `morning_brief` in solo) and the caller is not the principal. A request with no `x-caller-email` counts as the principal; with no principal on the roster, no signed-in caller does. This governs starting a run only. A run started by hand (this endpoint, or `run_workflow` / research in chat) records its starter as `owner_person_id` and is theirs alone: `GET /workflows/runs` lists the team's (scheduled, no owner) runs plus the caller's own and the ones waiting on them, and `GET` / `DELETE /workflows/runs/{run_id}` and `POST /workflows/runs/{run_id}/decision` are a 404 for anyone else's, the principal included. A signed-in email that is on nobody's People entry cannot start one (403), and a run whose starter can't be read is refused (503); its steps read and draft documents as the starter only. `POST /workflows/runs/{run_id}/decision` with `{decision: \"approve\"|\"reject\"}` answers a run awaiting a yes/no sign-off from the web app — 403 unless the caller is the person it waits on or the principal, 409 if it isn't awaiting or needs a written reply — through the same `apply_resolution` a chat reply uses. Custom create/update also rejects (422) any action-step tool that does not resolve, and so does `POST /workflows/custom/{name}/activate` when turning a workflow **on** — that click is the approval for a tool workflow chat saved switched off (turning off is never checked). Turning on a workflow with action steps requires the body `{is_active: true, definition}` carrying the definition the user reviewed; if the stored one differs (ignoring `is_active` and timestamps) it returns 409, and the switch itself is a compare-and-set on the validated revision (`dynamic_store.activate_if_unchanged`, keyed on `updated_at`), so a save that lands while the tool check awaits the gateway — even from another process — is a 409 too, so only what was shown gets switched on. `GET /workflows/custom` lists inactive definitions too; `GET /workflows/{name}` 404s for them. The two `tools` endpoints return `{tools: [{name, description, read_only, source}]}` for the advanced editor's tool picker and the review card; `read_only` is true only when the name says it reads (a label, not a control). The run stream adds a `progress` event (`step_id`, `summary`) while an action step uses tools. Each catalog entry (`WorkflowMeta`) carries `is_custom`, `background` (true for the system-run built-ins the UI files under “System”) and `quick` (true for a Quick workflow, one per playbook nothing else follows, named `quick_<playbook>`). The run endpoint streams SSE; its `awaiting_human` event carries `delivery` (`self` \\| `sent` \\| `suppressed` \\| `alerted` \\| `failed`) alongside `person_id` / `question` / `awaiting_until`, because the outbound guard or an unreachable approver means the question may never have been asked — a client must not render “waiting on them” for those. A paused run ends the stream with neither `done` nor `error`, so that `awaiting_human` frame carries `terminal: true` to say so explicitly, plus `resumable` — whether the run continues by itself once answered or stops at the gate. `GET /workflows/runs/{run_id}` returns the run row with the resume payload replaced by a small `resume_progress` summary (`gate_step_id`, `completed_step_ids`, or `null`): the payload holds every completed step's full text, and the run-detail page polls this endpoint while a run is unfinished. |\n| `knowledge` | `/knowledge/builtin`, `/external`, `/search` | MBA knowledge base and external sources. `/search` is the introspection parallel of `retriever.retrieve` and mirrors its filters — including the `general` catch-all on the company collection — so the panel cannot report a specialist seeing nothing that it does retrieve in chat. It accepts `general` in `domain_filter` for the same reason. `/search` applies the same review gate as `retrieve()` — withheld (pending or rejected) built-in, external and failure rows are dropped, so the panel never advertises knowledge the Executive cannot use; it degrades to unfiltered rather than failing if review state is unreadable. `POST /knowledge/builtin` refuses a name already held by shipped content, and `POST|DELETE /knowledge/failures` now register and deregister for review — without that a user-authored failure case study had no review row, so no SME decision could block it. |\n| `skills` | `/skills`, `/skills/search`, `/skills/{name}/restore` | Skill registry (the UI's **Playbooks**) CRUD and search. `PUT` on a built-in saves a company customization that shadows it; `DELETE` returns `{name, outcome}` — `deleted`, `reverted` (customization removed, built-in back) or `hidden` (built-in hidden for this company); `restore` un-hides; `GET /skills?include_hidden=true` lists hidden built-ins. List and detail carry `used_by` — `[{name, title, is_custom}]` of the workflows that follow the playbook (switched-off custom ones included) (`WorkflowMeta.playbooks` is the other direction). |\n| `evals` | `/evals/scenarios`, `/evals/runs` | Eval scenarios and run results. `POST /evals/runs` is a 403 when a workflow scenario names a workflow that only the principal may start in the mode it runs in (the weekly review; the morning brief in solo) and the caller is not the principal — the Jobs page's rule for starting a run; stored results are read under the existing `/evals/runs` rules. |\n| `memories` | `/memories/{decisions,initiatives,advice}`, `GET /memories/facts`, `POST /memories/facts/{fact_id}/{retire,approve,decline}`, `GET /memories/facts/approval`, `PUT /memories/facts/approval/{person_id}`, `GET /memories/people`, `GET /memories/people/{person_id}/conclusions` | Episodic-memory record management; the standing facts the Pulse Corrections tab reads (`{facts: [...], can_retire, retirable_ids, can_review}`; each fact carries `recorded_by_role` / `recorded_by_name`; `?include_inactive=` default true, honoured for the principal only; anyone else gets the active rows plus their own `proposed` ones, with the provenance (`session_id`, `turn_id`, `recorded_by_person_id`, `retired_reason`, and `source_quote` except on their own facts) blanked, never retired, replaced or declined text; `?limit=` 1–500) and their retire (the principal any fact, a teammate only one they recorded that the principal has not approved — else 403; the audit row names the teammate; 404 for an unknown or `profile` row, 409 when no longer active; writes a private `fact_retired` audit row); approve / decline a teammate's `proposed` fact (principal only, else 403; 409 when it is not proposed; private `fact_reviewed` row); and the per-teammate \"needs my approval\" switch (on by default; `GET` lists `[{person_id, full_name, needs_approval}]` for every active team member but the principal, `true` for anyone not yet set; `PUT` takes `{needs_approval}`, 404 for anyone else; both principal only; private `fact_approval_changed` row); plus the read-only peer-memory overview the Pulse People tab reads (`{status: ok|disabled|error, people: [...], conclusion_total}`; `?recent=` 1–50, default 5) and one person's full conclusion list, paged newest first (`{status, person_id, items, page, size, total, has_more}`; `?page=` ≥1, `?size=` 1–100, default 50; 404 when the person has no peer). Peer memory is each person's own: the overview holds only the caller's entry (none for a caller off the roster; `conclusion_total` counts only theirs), the principal included, and `/conclusions` for anyone but the caller is a 404. |\n| `review` | `/review/items`, `/review/stats`, `/review/trusted-defaults`, `POST /review/curate`, `POST /review/bulk-approve` | Knowledge review queue and annotations. Shipped knowledge registers as a trusted default (`approved` with a NULL `reviewed_at`), so the queue starts empty; `curate` (`{domain, action: start|stop}`, domain required) opts one domain in or out, and `trusted-defaults` returns domain → unreviewed count. `bulk-approve` takes a non-empty `domain`, `item_ids` (max 500) or `all_pending: true`, and 400s with no selector; it returns `{approved_count, item_ids}`. |\n| `onboarding` | `/onboard/interview/understand`, `/onboard/interview/start`, `/onboard/interview/message`, `/onboard/interview/draft`, `/onboard/interview/{id}`, `/onboard/interview/commit` | Conversational company setup. The `onboarding_interviewer` agent asks clarifying questions, then emits a draft profile + people + departments. **Nothing is persisted until `POST /onboard/interview/commit`** — the interview itself never writes. `start` is multipart (description + optional deck/one-pager uploads). `commit` answers 403 when the draft would make a different person the owner and the caller is not the current owner (setup demotes the old principal, and any signed-in user can open it). It also takes an optional `owner_email` — the principal's sign-in email, which the review screen pre-fills from the signed-in user — checked before anything is written (422 if it is malformed, already another person's, or would replace the principal's existing email — setup only fills in a missing one; 403 if the caller is not the owner and it is not the address they signed in with, since the roster is the sign-in allow-list) and saved on the principal's People row, so the owner's chats resolve to them from the first turn. |\n| `workflow_designer` | `POST /workflows/designer/start`, `POST /workflows/designer/edit`, `POST /workflows/designer/message`, `POST /workflows/designer/draft`, `GET /workflows/designer/{session_id}` | The conversational New workflow wizard, and editing a saved workflow by conversation. `/edit` (`{name}`) opens a session on a saved custom workflow with no model call: the stored definition is the starting draft, and every response then carries `editing` (its name) and `original` (the stored definition as the session opened on it); drafts keep that name and its on/off state whatever the model writes (404 for an unknown workflow). The UI saves one with `POST /workflows/custom/{name}/save-edit` (`{session_id, definition}`): the server saves the session's own draft, never a client-sent body, and only when it equals the `definition` the user reviewed (a later message is a 409, so the click saves only what was shown); it re-validates it, keeps the stored on/off state, and writes only if the stored workflow is still the version the conversation opened on — otherwise 409, as it is for a session about another workflow or one with nothing drafted yet. It sits under `/workflows/custom` so anything that limits writes to custom workflows covers it too. The `workflow_designer` agent asks clarifying questions (`phase: \"question\"`, with `hint` and up to four suggested `options`), then returns `phase: \"draft\"` with `draft.definition` (exactly the body `POST /workflows/custom` takes), `summary`, and `assumptions`; every response carries the `transcript`. **Nothing is persisted here** — sessions are in memory, and the UI saves through `POST /workflows/custom`. Errors are fixed strings (422 over the 8,000-char message cap or the transcript cap, 404 for an expired session, 409 while another turn on the same session is still running, 502/504 for model failure or timeout). Mounted ahead of the `workflows` router. |\n| `python_jobs` | `GET /python-jobs/{job_id}/{name}` | A result file a Python job (`run_python_job`) wrote, the principal's alone (403 otherwise), served as a download (`Content-Disposition: attachment`, `Content-Security-Policy: sandbox`, `nosniff`, `Cache-Control: no-store`); 404 for a bad id or name or a file past its 7 days. |\n| `saved_tools` | `GET /saved-tools`, `GET\\|PUT /saved-tools/{name}`, `POST /saved-tools/{name}/rollback` | The scripts the Executive saved as tools (Settings → Advanced → Saved tools), the owner's alone (403 otherwise). The list gives each tool's name, description, on/off, current version, `kind` (`script` or `python`, a kept Python job) and the tools it calls, plus whether `SAVED_TOOLS_ENABLED` is on; the detail adds the script, every version and the last 20 runs. `PUT {enabled?, workflows?, version?}` turns one on or off (off also turns it off for workflows, so turning it back on never re-arms unattended runs), and turns the named `version` — the one the owner looked at, never whatever is current by then — on for workflows (or workflows off); a Python tool never turns on for workflows (409) — turning workflows on lets more run unattended, so it needs a request tied to the owner (signed sign-ins or local login, as Take the lead's switches do; 409 otherwise); `rollback {version}` makes that version the one that runs. Both are audited as `saved_tool_changed`. The Executive saves tools itself through `run_script`; there is no create route. |\n| `onboarding` (fallback) | `/onboard/start`, `/onboard/answer`, `/onboard/status/{id}` | The original 12-step form wizard, kept as a fallback (UI at `/onboard?mode=form`) and used by the `openexecutive onboard` CLI. Its final answer saves the profile and adds people with their emails and a principal, so once there is an owner only they can send it (403, and the session rolls back so the owner can still finish it). `/onboard/start` reads the workspace mode once for the session: solo skips the Team Size, Team Members and Fractional Executives steps. Each status response's `current_step` / `total_steps` count only the steps that session asks, and `optional` says whether the current step can be skipped. |\n| `company-profile` | `GET\\|PATCH /company-profile` | Structured company profile. |\n| `alerts` (private cards) | `POST /alerts/{id}/ack`, `/reopen`, `/alerts/bulk-ack` | An alert tagged `private:principal` (see Today) is a 404 on ack and reopen for anyone but the principal, and their bulk sweep skips it. A roster-request card (source `roster_request`) is a 409 on `ack` even for the principal — it is answered at `/people/requests/{id}`, since acking would leave the sender waiting with nothing to show it — and, as an exempt source, `bulk-ack` and `reopen` skip it. |\n| `auth` | `GET /auth/allowed-emails` | Non-archived **team members** with an email, offered to the UI's sign-in gate — a contact is never on it, so adding a client to reach them by email no longer lets them sign in. The gate takes the **union** of this roster and its own `ALLOWED_EMAILS` env list, so env entries always work (and are checked first, short-circuiting the fetch) and roster entries are additive — a roster change can never revoke a configured operator. With local login (`make dev` with Google sign-in not set up — the UI then listens on 127.0.0.1 only) the UI signs in without an email, sends no `x-caller-email`, and the backend resolves the caller to the principal, as it does for the CLI. |\n| `fixtures` | `/fixtures`, `/fixtures/{name}/load` | Demo-company fixture load/reset/snapshot. `POST /fixtures/reset` leaves no departments, like a new install (`departments_seeded` is always 0). |\n| `clients` | `/clients`, `/clients/generate`, `/clients/cockpit`, `/clients/save`, `/clients/{slug}/activate` | Client-company slots — multi-client mode for fractional use; create from current/blank/AI-generated intake draft (pasted notes and/or file attachments, multipart on `/clients/generate`), save-back switching, engagement metadata (PATCH), cross-client cockpit, delete. See the Client Companies section. |\n| webhooks | `POST /webhook/telegram`, `POST /webhook/google-chat`, `POST /webhook/slack` | Inbound integration callbacks (self-verified). `/webhook/slack` answers 503 unless `SLACK_SIGNING_SECRET` is set. |\n| MCP server | `POST\\|GET /mcp` | Open Executive exposed as an MCP server (Streamable-HTTP). Not a router — a mounted FastMCP sub-app. Same shared-secret gate (`x-api-key`); exposes company context as resources and the specialist council as tools. See the MCP Server section. |\n\n**Take the lead** (`api/routes/take_the_lead.py`): `GET /take-the-lead` returns the As the Executive switch, its six Ask first switches (`{kind, label, hint, on}`, the hint a plain line on what each one catches), the company's rules, whether this server can tie a change to the owner (`available`), and whether the Executive is paused. `PUT /take-the-lead` takes `{enabled?, ask_first?}`, and `POST /take-the-lead/rules` / `DELETE /take-the-lead/rules/{id}` add or remove a company rule (`{kind: person|domain|words|amount, value}`, 422 on a bad one). All four are the principal's alone (403). Anything that lets more happen on its own (turning it on, turning an Ask first off, removing a rule) needs a caller the API can tie to them (`delegation.verified.caller_refusal`, else 409). `GET /today/activity` adds `took_the_lead` rows (what Take the lead did, named, not quoted) and, for the caller alone, `sent_as_you` rows for what Handle it for me sent as them. Under `/delegation`: `PUT /delegation/take-the-lead {enabled}` is Take the lead as you (the owner's only for now; turning it on turns Handle it for me on), and `GET|POST /delegation/take-the-lead/rules` and `DELETE …/rules/{id}` manage the caller's own rules. `HandleItOut` gains `lead` and `lead_available`. A `take_the_lead_action` decision is approved or rejected through the existing `/decisions/{id}/approve|reject`. Training: the response also carries `training` (In training: everything waits unless allowed), `learned` (`{id, label, feature, example, uses, created_at}`: what it was allowed, the edit kept as an example by field) and `suggested` (`{key, label, approvals}`); `PUT` takes `training?` too, and leaving training while on needs the same provable owner as turning it on. `POST /take-the-lead/learned {key}` allows a current suggestion (404 for anything else; provable owner), `DELETE /take-the-lead/learned/{id}` makes it ask first again. `learned` also lists the owner's own Act as me senders (`feature` `act_as_me`), never anyone else's, and Remove works on those too. On a `take_the_lead_action` card, `POST /decisions/{id}/approve` takes `edits.input` (the card's editable fields, 422 for any other; the principal's alone, 403 for an area approver) and `edits.allow: true` (a training card only, 422 otherwise; the principal from a provable request, 403 otherwise), finishing `approved_with_edit` when a field changed.",
+ "markdown": "The FastAPI surface lives in `openexecutive.api.main`, which mounts one router per topic. A shared-secret middleware gates every request via an `x-api-key` header, except `/health` and the `/webhook/*` callbacks (which carry their own verification). With local login (`OE_LOCAL_LOGIN=1`, which only `make dev` sets, and only while neither Google nor SSO sign-in is set up) a second middleware answers 403 to any request whose raw `Host` is not `localhost`, `127.0.0.1` or `[::1]` (except a webhook that verifies its caller: Google Chat and Slack always, Telegram only with a `TELEGRAM_WEBHOOK_SECRET` Telegram can send), and to any write whose `Sec-Fetch-Site` is present and not `same-origin`/`none`. The UI signs that owner in with no password and this API usually has no secret, so without it a web page in the owner's browser could drive the API as the principal — by DNS-rebinding its own hostname to 127.0.0.1, or by posting a plain form from another site or localhost port. CORS is restricted to configured UI origins. The gate is off when `BACKEND_SHARED_SECRET` is unset — intended for local development only — so any internet-reachable deployment sets `OE_PUBLIC_DEPLOYMENT`, and `create_app()` then refuses to boot without the secret rather than serving an unauthenticated API.\n\n**Who is calling** is read in one place, `api/caller.py`, which answers with a `Caller` kind and email; a test fails on any other read of an `x-caller-*` header. By default (`open`) it is the `x-caller-email` the UI proxy stamps from the sign-in, trusted because whoever holds the shared secret is trusted as the proxy, and a request with none runs as the principal. **Signed callers** remove that trust. With `CALLER_ASSERTION_PUBLIC_KEYS` set, the proxy signs every request with an Ed25519 key only the UI holds, and a third middleware checks the `x-caller-assertion` header:\n\n- **What it covers.** It is good for one method, the exact path and query, at most 120 seconds, and one use. The API keeps used ids in memory.\n- **What it names.** A `user`, by email, or the `operator`: local login, which runs as the principal.\n- **No caller header at all** is a `service`, for scripts or MCP clients holding only the secret. It is never the principal, not even on an install with no owner yet, so it can't save a first setup.\n- **Anything else is a 401.** That includes an `x-caller-email` without an assertion (`caller_assertion_required`), and a bad, expired, reused or mismatched assertion (`caller_assertion_invalid`).\n\nPublic keys that can't be read stop the boot, rather than fall back to trusting the header. `scripts/make-caller-keys.py` makes the pair, and `scripts/mint-caller-assertion.py` signs one request for curl. The API holds only public keys. Setup's API-protection check warns while a shared secret is set without them.\n\nEndpoints grouped by router:\n\n| Router | Key endpoints | Purpose |\n|---|---|---|\n| `chat` | `POST /chat`, `POST /chat/stop`, `POST /chat/add`, `POST /chat/background` | SSE stream of a live Executive turn. `POST /chat/background` starts a new chat that the server answers to the end with nobody connected (a phone approving a card, then closing the app): it answers `202` with the new `session_id` once the same caller, access and quota checks as `/chat` pass, the turn saves itself as a `/chat` turn does, and one caller may have three running at once (`429` beyond that). `POST /chat/add` hands the running turn a message the person typed while it works; the turn takes it at its next step and the stream names it in a `message_added` event, and a second `POST /chat` on the same conversation waits for the first to finish. The request body optionally carries `page_context` (route, page title, guide section id, and a form descriptor with current values) — sent only by the Ask OE side panel so the Executive can explain the current page or propose form values. When the Executive calls its `propose_form_values` tool, the stream emits a `form_patch` event (`{type, form_id, fields, rationale, iteration}`) that the panel applies into the on-screen form as highlighted suggestions; the user reviews and saves through the form's normal submit path. Each round of tool calls also emits an `activity` event (`{type, label, tool, iteration}`) naming what is running — progress only, forwarded verbatim and never persisted with the message, unlike an `action_taken` chip. A `session_id` in the body must be a chat the caller may use (the same ownership rule as `sessions`). If it isn't, the turn runs in a fresh chat instead, so nothing is written into someone else's session and the reply looks the same as for an unknown id. An unknown namespaced id (`slack:…`, `telegram:…` and so on) is ignored and a fresh chat is started, because those ids belong to the channel adapters and claiming one first would make the caller its owner. The client mints a `client_turn_id` and sends it with the turn; `POST /chat/stop` takes that id and halts the turn mid-flight, so the button is live from the moment Send is pressed rather than only once the first SSE byte arrives. A stopped turn emits a `stopped` event and then the usual `done` over the same still-open stream — never an `error` — and whatever streamed so far is persisted and flagged. The endpoint returns 404 both for an unknown id and for another caller's turn, so it cannot be used to probe which turns are live. An optional `memory_text` (1–2000 chars) is what peer memory records as the caller's words instead of `message` — the briefing handoffs send it because their seed quotes the Executive's own card; `POST /chat/upload` derives it from the typed text and filenames. Since it never reaches the transcript, the `chat_turn` audit row keeps it beside `message`. After the reply — a stopped or timed-out one too, before the `stopped` / `error` / `done` frames — a turn that looked at anything sends one `sources` event, `{type, session_id, sources: [{kind, title, url}], unavailable: [area]}`: `kind` is `company`, `knowledge`, `notion`, `research`, `document` or `web`, `url` is only ever an http(s) link or, for an earlier document, `/artifacts/<id>`, at most 12 sources (6 from the web, pages the reply cites first), and `unavailable` names parts of the analysis (\"finance\") whose specialist failed or said nothing every time it was asked. It is persisted with the reply as `chat_messages.sources`. A request that carries `x-caller-email` marks its turn signed in (`Session.web_caller_signed_in`), which Act as me needs on a deployment without local login. |\n| `sessions` | `GET /sessions`, `GET /sessions/{id}`, `GET /sessions/{id}/messages`, `DELETE /sessions/{id}`, `POST /sessions/{id}/messages/{message_id}/feedback`, `GET /sessions/{id}/followup` | List, read and delete saved chat sessions (assistant messages carry their row `id` and any `feedback`). The list is scoped to the caller, and each chat carries `preview`: its last message as one plain line, Markdown marks stripped and cut at a word near 120 characters, with `You:` before the person's own words. With `?q=` it is a search: the caller's own chats whose title or message text contains the words (case-insensitive, `LIKE` with wildcards escaped), each with `match_count` (messages that contain them; 0 for a title-only match) and `snippet` (the newest matching message, around the match, prefixed `You:` or `Executive:`). Scoped exactly like the list, so nobody else's conversation can match. The Chats page searches this way and opens a result with `?find=`, so the chat starts on that message and steps through the others. A chat-app conversation started with /new is left off the list until its first turn (`memory.conversations`). Every per-session route checks ownership, because session ids are guessable (`slack:dm:<user id>`, `telegram:<chat id>`): only the session's owner or the principal gets through, and a legacy row with no owner is the principal's alone. A teammate's own chat (a web chat, a DM, a private Telegram chat; not a channel or group conversation, `chat._shared_space`) is theirs alone, the principal included, as their mail with the Executive is. Anyone else gets the same 404 as for an unknown id, so the routes can't be used to find out which chats exist. The one other way in is having started the chat in this process (matched by Person id or verified email, so it survives the caller being added to the roster), which is how a caller the roster can't resolve keeps using their own chat. That record is in memory, so after a restart such an ownerless chat is refused for reads. The principal can continue someone else's ownerless chat without becoming its owner. Deleting a session also drops its in-memory copy, so a later turn naming that id starts fresh. The feedback route records 👍/👎 (`{\"feedback\": \"up\"|\"down\"|null}`) on one assistant reply; only the session's own caller or the principal may rate it. The rater is stored (`feedback_by_person_id`), and a 👎 on a reply to the rater's own message triggers a working-style pass for them (paced; see attunement). The chat stream's `done` event carries `message_id` for the persisted reply. The followup route returns `{\"suggestion\": string|null}` — one suggested next user message, written by the utility-fast model from the last few turns, that the chat composer shows as its greyed placeholder (Tab / → or the inline chip fills it in without sending). It is null unless the session ends on a persisted assistant reply, and null (uncached) when the model fails. Same ownership check as feedback; successful suggestions are cached in-process per (session, reply id). Messages from `GET /sessions/{id}/messages` carry each assistant reply's saved `sources` (the same shape as the chat stream's `sources` event, without `type`), so a reloaded chat shows them again. |\n| `today` | `GET /today`, `GET /today/activity`, `GET /today/activity/daily`, `GET /today/brief-delivery`, `GET /today/top-three`, `GET /today/weekly-review` | Morning-brief aggregation, recent-action rail, and Pulse heartbeat heatmap (per-day activity counts). `GET /today` carries the viewer's cached \"What's going on\" `narrative` with `narrative_generated_at` (when it was written) and `narrative_stale` (a fresher one is being written — the page re-polls; see Today). A proposal backed by a roster request carries `roster_request` (the new sender's server-derived details and held-message previews; the principal's only) and is answered through the `people` routes below. `brief-delivery` returns the latest morning brief or end-of-day digest that didn't reach the owner, as `{brief, at, problem, fix, readable}` in plain language (`readable` is false when it couldn't be written, so there is nothing to read), or null — null while briefs go out, once the cause is fixed, while the scheduler is off, and for anyone but the owner (`chat._caller_is_principal_or_unclaimed`); it carries no address or channel id. `top-three` and `weekly-review` feed the solo Briefing's own cards and answer only in solo mode and only to the principal (`workflows._caller_is_the_principal`: a principal match or no caller header, never an unclaimed roster); anyone else, and team mode, get null. `top-three` returns `{items: [{key, kind, text, why, slot}]}` — the morning brief's top three from the same code (`build_top_three(principal_due_soon())`), `slot` a free block like `10:00–11:00`, `\"\"` when none is left, null when no calendar was read — and is kept off `GET /today` because its calendar read may take up to 4 s. `weekly-review` returns the latest completed `weekly_review` run as `{run_id, completed_at, period, top_three, excerpt}` (next week's top three as plain text; `excerpt` when that list is empty), or null when none has completed. |\n| `people` | `GET\\|POST\\|PATCH /people`, `POST /people/{id}/archive`, `GET /people/requests`, `GET /people/requests/{id}`, `POST /people/requests/{id}/approve`, `POST /people/requests/{id}/decline`, `GET\\|POST /people/{id}/open-loops`, `POST /open-loops/{loop_id}/close`, `GET /people/{id}/outreach`, `GET\\|PUT\\|DELETE /people/{id}/attunement` | Roster CRUD and scope lookups. Every `Person` carries `kind` (`team` / `contact`). `GET /people` lists team members only unless `?include_contacts=true` (the People page asks for both; the department-head and workflow pickers keep the default); `POST` / `PATCH` accept `kind`, and a principal contact is a 422 either way. They also take `email_aliases` (other addresses the person writes from, at most 10; on `PATCH` the full list replaces the current one, omitted leaves it): an alias matches their mail and may be emailed but never signs in; a malformed one is a 422, and an alias — or a new primary `email` — that is already someone else's address or alias is a 409. Every `POST` / `PATCH` then closes any pending roster request whose sender now matches someone and replays its held messages. Adding, editing and archiving (`POST`, `PATCH`, `/archive`) are the principal's only, or anyone's while no one is principal yet so a first setup can add its owner (`chat._caller_is_principal_or_unclaimed`, the check behind `PUT /workspace`; a request with no `x-caller-email`, such as the CLI or local login, counts as the principal). Anyone else, a person editing their own row included, gets 403 before the person lookup, so it can't probe which ids exist. Contacts are private to the principal (`caller_is_principal`: the signed-in email is the principal's, or no `x-caller-email`; stricter than the owner rule above, so while no one is principal nobody is): for anyone else `include_contacts` is ignored, a contact's id is a 404 on `GET /people/{id}` and open-loops (exactly as for a missing id; `PATCH` and archive are already a 403 for them), and creating a contact or turning someone into one is a 403, even on a first setup with no principal yet. `GET /people/me` returns `{person_id, is_principal}` so the UI offers contacts to the principal only. `GET /people/by-scope/{token}` never returns a contact. Archiving a person closes their open loops. The open-loop routes list what a person owes (soonest due first) and close one (`reason`: done / not_needed / cancelled) — both the principal's or the owner's only (403 otherwise, via `people.store.is_principal_or_self`). `POST /people/{id}/open-loops` `{task, due_date?}` assigns that person a task (201 with the loop; `due_date` is a local date, due 17:00 in the user's timezone, clamped to 60 days, default 2 days): any active team member may assign anyone on the team, 403 for anyone else; 409 for an archived person, a contact (as seen by the principal), a person at the open-loop cap, a duplicate or open loops turned off; 422 for an empty task. Nobody is messaged at assignment — the owner is chased once it is due. `/outreach` returns, per kind of proactive outreach over 30 days, `sent` / `replied` / `acted` / `ignored` / `pending` (voided DMs are excluded) — same access rule, checked before the person lookup so it can't probe which ids exist. `/attunement` reads, replaces (`rules`: up to 4 strings, or omitted to only set `locked`) or resets the person's working-style rules — same access rule, 404 for an archived person; `PUT` rejects a rule that is not purely about style with 422. Archiving a person also drops their working style and frees their aliases. The **roster-request** routes (declared before `/people/{person_id}`) are the principal's alone (`caller_is_principal`; anyone else gets 404, as for a missing id): `GET /people/requests?status=pending|approved|linked|declined|expired|superseded|all` lists up to 100 (`id`, `channel`, `channel_ref`, sanitised `display_name`, `profile_email`, `on_company_domain`, `suggested_kind`, `suggested_person_id` / `_name`, `status`, `resolved_person_id` / `_kind`, `message_count`, timestamps, `ack_sent`, and `previews` of the held messages while pending — shown to the principal, never to a model); `approve` takes either `link_person_id` (+ `replace_channel_id`) or `full_name` + `kind` (no default: 422 without both) + optional `role`; `decline` takes no body. An answered request is a 409, a chat id or address already on someone else is a 409; the held messages are replayed (or dropped) in the background. |\n| `departments` | `GET\\|POST\\|PATCH\\|DELETE /departments`, `.../goals` | Department state and per-department goals. `PATCH /departments/{slug}` also takes `watched_entities` (≤50 names, `[]` clears) — the entities the research watch policy may add watches for and route to this department. A `cadences` spec that does not parse is rejected with a 422 (empty means no cadence). A new `head_person_id` must be a team member: an unknown id and one of the principal's contacts get the same name-free 422 (`person_id N is not on the team.`), so the route — open to any signed-in user — never reveals that a contact exists; an unchanged head is not re-checked. `POST .../goals` requires only `key_result`: a missing or blank `period_value` is filled with the current period for `period_type` in the user's zone (the same label `create_goal` uses), and `target` defaults to empty. `PATCH .../goals/{id}` rejects a blank `key_result` or `period_value` with a 422; `target` may be cleared. |\n| `agents` | `GET /agents`, `GET /agents/models`, `GET /agents/models/options`, `GET /agents/presets`, `POST /agents/presets/{id}`, `PATCH /agents/{id}`, `POST /agents/{id}/test` | Specialist metadata and Council overrides. `/agents/models` is the flat model allowlist the PATCH validator uses; `/agents/models/options` returns the same ids in the same order, each tagged with `provider`, `provider_label`, `route` (`direct` / `openrouter` / `local`) and a display `label`, for the Council's Provider → Model picker. `AgentDetail` carries `instructions` (additional instructions, appended after `prompt`, which stays the built-in prompt or its replacement); `PATCH` and `/test` accept `instructions` (max 4,000 chars, blank clears it). `GET /agents/presets` lists the Fast, Balanced and Thorough quality presets (each with `available` and the resolved `model` for this install) plus `active` (null when the council is Custom), `base` and `custom_agents`; `POST /agents/presets/{id}` applies one by writing per-agent model and deep-reasoning overrides and returns the same shape (404 unknown preset, 409 when the install allows none of its models). Each `AgentMeta` carries `visibility` (`core` or `internal`) so the Council's simple view can list core agents first. |\n| `personas` | `GET\\|PUT\\|POST\\|DELETE /personas` | Executive persona definitions. |\n| `alerts` | `POST /alerts/{id}/ack`, `POST /alerts/bulk-ack`, `POST /alerts/{id}/reopen`, `POST /alerts/review` | Groom the briefing queue. `ack` (`read`/`ack`/`dismissed`, optional `mute_topic`) also teaches a watch-sourced alert's watch (`trust_score`); `bulk-ack` takes explicit `alert_ids` and/or `older_than_days` + `category` (exempt sources skipped); `reopen` is the Undo for a `resolved` / `expired` / `dismissed` alert (never `ack` or an artifact / decision row; back to `unread` with every review stamp cleared and its TTL anchor refreshed, so it is live again even if older than its TTL); `review` runs the Executive's relevance review on demand (ignores the interval; still honours `ALERT_REVIEW_ENABLED`, the per-pass caps, the single-flight lock and per-alert idempotency) and returns the move counts. `bulk-ack` caps ids at 500, needs `older_than_days >= 1`, and applies watch feedback per row; `mute_topic` must be one of the alert's tags or a scoped `external:` / `department:` pattern. Alerts still surface only through the briefing (`/today` reads the live view) — there is no list, prefs, mutes, or live-stream endpoint. |\n| `artifacts` | `GET /artifacts`, `GET /artifacts/{id}`, `GET /artifacts/{id}/download?as=`, `POST /artifacts/{id}/archive`, `POST /artifacts/{id}/restore`, `DELETE /artifacts/{id}` | Browsable deliverables in any format — Markdown, HTML, Word, Excel, or a link into a connected app — with `format`, `downloads`, `external_url`, `supersedes_id` and, for an extension's kind of document, `collection` / `collection_label` on every summary (`orchestrator.extensions`; archive, restore and delete tell the collection first and answer 502 if it refuses). `download` renders the file (Word/Excel built at request time from the stored text; any Markdown artifact, workflow output included, can export `?as=docx`) and is always an attachment with `nosniff` and a sandbox CSP. Every route answers for the signed-in caller (`artifact_records.Viewer`): a drafted artifact is its owner's alone (`alerts.owner_person_id`, the person whose conversation published it; none = the principal's), a workflow output follows its run (`persistence.run_visible_to`), and anything the caller may not see is a 404 like a missing id, the principal included. `POST /alerts/{id}/ack` and `/reopen` answer 404 for someone else's drafted artifact the same way. |\n| `decisions` | `GET /decisions`, `POST /decisions/{id}/{approve,reject,cancel}`, `GET`/`PUT /decisions/classes/meeting_scheduling`, `GET /audit/reliability` | Approval-gated decision instances (e.g. calendar bookings). What approve/reject/cancel do is per class (`DECISION_CLASSES`, one `DecisionClassSpec` each; meeting booking books on approve and deletes on cancel; `delegation_reply`, an inbox reply card, is its approver's alone (`approver_only`) — hidden from everyone else, the principal included when it is a team member's — sends its Gmail draft on approve and, after a reject, deletes the draft unless it was edited), and an instance of a class with no spec gets 409 instead of being carried out as another. Only the principal, or the person the decision went to for approval (`approver_person_id`), may approve, reject or cancel it. Anyone else gets 403, and a decision nobody was asked about, or one private to the principal, is the principal's alone. The resolver is recorded as `resolver_person_id`, and a reject that loses the race to another answer is a 409. `GET /audit/reliability` is the principal's alone (403 otherwise), because it counts private decisions too. A proposed booking is surfaced on the briefing as a companion alert; approve/reject/cancel here also clear that alert. An approval whose caller resolves to the principal runs the booking inside `grant_contact_egress()`, so a meeting the principal proposed with a contact can be created. Such a booking's payload is marked `private`: it is proposed to the principal, left out of anyone else's `GET /decisions`, and a 404 for them on `GET /decisions/{id}` and every action route. Reviewed in the briefing — there is no separate proposals page. The class routes read and set how meetings are handled: both return `{decision_class: \"meeting_scheduling\", mode: \"propose\" \\| \"auto_execute\"}` (default `propose`); the `PUT` body is `{mode}`, principal-only (403 otherwise, open before a principal exists — but then only `propose` can be set: `auto_execute` is a 409 until a non-archived principal is on the roster), 422 on any other mode, and a change writes a `decision_class_mode_changed` audit row. In solo mode the class mode decides a booking, and `auto_execute` applies only to a turn the principal started on a verified surface — any other turn gets a proposal. |\n| `delegation` | `GET`/`PUT /delegation`, `GET`/`PUT`/`DELETE /delegation/voice`, `POST /delegation/voice/learn`, `POST /delegation/voice/describe`, `POST /delegation/voice/signature`, `PUT /delegation/inbox`, `POST /delegation/inbox/check`, `GET /delegation/replies`, `PUT /delegation/replies/{id}/text`, `GET /delegation/actions`, `PUT /delegation/handle-it`, `PUT /delegation/disclosure`, `PUT /delegation/training`, `DELETE /delegation/learned/{id}`, `GET /delegation/handled`, `PUT /delegation/team` | Act as me — the **caller's own** delegation only; nobody sets it for anyone else. It answers the principal, and a team member once the owner lets them; anyone else gets 403 `not_available_yet`. `GET /delegation` returns `{enabled, gmail: {status, message, email, connect_command, outlook_connect_command, provider}, inbox: {enabled, status, message, watch_since, last_poll_at, checking}, handle_it: {enabled, mode, available, sent_today, lead, lead_available}, training: {replies, follow_ups, actions, drafts, learned}, disclosure: {enabled, handle_it, approved_sends, line}, team}`, where `team` is the owner's switch for the principal while `DELEGATION_TEAM_MEMBERS` allows it — `{enabled, members: [{person_id, name, enabled, inbox, drafts_30d, sent_30d}]}`, counts only — and `null` otherwise. `PUT /delegation/handle-it` takes `{enabled?, mode?}` (`mode` is `careful`, `balanced` or `bold`; anything else is a 422 `bad_mode`) from a caller the API knows is that person — 403 `not_yours`, 409 `caller_signing_required` with only the shared secret — and turning it on needs Act as me and Draft replies to my inbox on (409 `act_as_me_off` / `inbox_off`); `GET /delegation/handled` returns `{replies: [{decision_id, sent_at, to_name, to_email, subject, body, open_questions, gmail_link, source}]}`, the caller's replies and follow-ups (`source` `follow_up`) sent on its own in the last 7 days; a card on `GET /delegation/replies` carries `source` the same way, `can_allow` while Replies (Follow-ups, on a follow-up) is in training, and `learns_style` while Drafts is. `PUT /delegation/disclosure` takes `{handle_it?, approved_sends?}` (any other field is a 422), the caller's own switches for the AI line on replies sent as them (`disclosure.enabled` is the install's `AI_DISCLOSURE`); turning one on needs no signed sign-in, turning one off does (409 `caller_signing_required`), and a change is audited as `delegation_disclosure_changed`. `PUT /delegation/training` takes `{replies?, follow_ups?, actions?, drafts?}` (any other field is a 422) and returns the `GET /delegation` shape; putting Replies or Follow-ups in, or Suggested actions out, needs no signed sign-in, anything else does (409 `caller_signing_required`), and putting Replies or Follow-ups in turns Take the lead as you off (`PUT /delegation/take-the-lead` on answers 409 `in_training` while they are). `training.learned` is `[{id, label, setting, example, uses, created_at}]`, what the caller allowed or kept in training (`setting` `replies`, `follow_ups`, `actions` or `drafts`; the kept example, or \"\"); `DELETE /delegation/learned/{id}` removes one of the caller's own (404 for anyone else's). `PUT /delegation/team` takes `{enabled}` from the principal alone (403 `principal_only`; 409 `not_available` when the install doesn't allow it). The Gmail `status` is `connected`, `not_configured`, `needs_reconnect`, `mismatch`, `no_email`, `shared_mailbox` or `error`, checked live with Google on each call. `PUT` takes `{enabled}` (any other field is a 422); turning it on needs the caller's own Gmail connected, otherwise a 409 whose code says why (`gmail_not_connected`, `gmail_needs_reconnect`, `gmail_mismatch`, `no_email`, `shared_mailbox`, `gmail_error`). The voice routes show, learn (from the caller's sent mail, one pass at a time — a 409 `in_progress`, `locked`, `too_soon`, `not_enough_mail`, `no_profile` or `changed` when it was locked or edited while learning, or a Gmail code), write it from the caller's own description (`POST /delegation/voice/describe` with `{description}`: one model call applies their words to the profile they have and returns it with a short `sample_reply`, saving nothing, so they review it and save it with `PUT`; a 409 `empty`, `too_long`, `in_progress`, `too_many` past 10 an hour or `no_profile`; no mailbox needed), edit and lock, and reset \"How I write\": a typed field that fails validation is a 422 `invalid_voice` listing each `rejected` field, and `clear_exemplars` / `clear_signature` drop those. `POST /delegation/voice/signature` reads the signature from the caller's Gmail settings again and changes nothing else (a locked profile stays locked), with the same Gmail codes as learning: a 409 when their mailbox isn't connected, it refuses the sign-in or the check itself fails (`gmail_error`), and a 502 `gmail_error` when reading the signature fails. On an Outlook mailbox it is a 409 `signature_unavailable`, since Outlook gives apps no signature to read, and the stored one is left alone. A caller `can_delegate` refuses — anyone but the principal, for now — gets 403 `not_available_yet`, which hides the Settings card. A request with no `x-caller-email` is a 403 `sign_in_required` unless the API runs under local login — a header-less call is not a sign-in (the caller is read through `api/caller.py`, as on every principal-only route). `PUT /delegation/inbox` takes `{enabled}` and turns **Draft replies to my inbox** on or off, returning the same shape as `GET /delegation`: on needs Act as me on (409 `act_as_me_off`) and the mailbox connected (the Gmail codes above, which keep their names for an Outlook mailbox too), and turning Act as me off turns it off too. `POST /delegation/inbox/check` starts a scan now in the background and answers 202 with the `inbox` state (409 `inbox_off` or `in_progress`). `GET /delegation/replies` returns `{cards: [...]}`, the caller's open reply cards from the database alone (no Gmail call): the decision id, the sender's name, address, relation and whether it was verified, the subject, when it arrived, what they wrote, the draft's recipients, subject and body, `open_questions`, `flags` and a Gmail link built from a fixed prefix. **Send** is `POST /decisions/{id}/approve`: it sends the card's draft exactly as it is in Gmail, after checking again that the caller is its owner — which on a server needs signed callers (409 `caller_signing_required` otherwise) — that both switches are on (`inbox_off`), no client slot is active (`client_slot`), the Gmail is connected (the Gmail codes), the draft is still there and unanswered (`draft_gone`, `you_replied`) and its recipients are allowed (`no_recipients`, `too_many_recipients`, `executive_recipient`, `not_from_you`). Recipients that changed, or a newer message in the thread, answer 409 `{code: \"confirm\", message, reasons, recipients}`; approving again with `edits: {recipients: <the list shown>, thread_moved_on: true}` is the second yes. **Send + allow** adds `edits.allow: true`: 422 `cant_allow` unless Replies (Follow-ups, on a follow-up card) is in training, 409 `learned_full` when their list is full, both before anything is sent; `edits.example: true` keeps a draft edited in the mailbox as how they write to that person while Drafts is in training. A draft edited in Gmail between those checks and the send is not sent (`draft_changed`). A double tap sends once (`already_handled`); Gmail refusing answers 429 `rate_limited` or 502 `gmail_error` with nothing sent, and a send Gmail didn't confirm answers 502 `send_unconfirmed` and is settled by the next inbox scan. Dismiss is `POST /decisions/{id}/reject`. `GET /delegation/actions` returns `{cards: [...]}`, the caller's own approval cards (`delegation_actions`): the decision id, when it was made, a one-line `why`, and each action's `index`, `kind` (`invite`, `message`, `add_contact`), `summary` and `text` (a message as it will be sent, or a meeting's description). **Approve** is `POST /decisions/{id}/approve` with `edits: {only: [indexes]}` (and `allow: true` for Approve + allow while Suggested actions is in training: 422 `cant_allow` otherwise or when nothing on it can be allowed, 409 `learned_full`; a card carries `can_allow`); it needs the card's own person signed in (403 `not_yours`, 409 `caller_signing_required`), Act as me on (409 `act_as_me_off`) and a card under a week old (409 `expired`), runs once (409 `already_handled`), and returns the decision with `final_payload_json.results`: per action `done`, `waiting` (sent to whoever approves meetings), `failed` with why, or `skipped`. Every change writes a private audit row. |\n| `history` | `GET /memories/history` (`?q=`), `PUT /memories/history/settings`, `PATCH`/`DELETE /memories/history/{id}`, `POST /memories/history/{id}/forget-conversation` | Always in the loop: the **caller's own** notes and switches only, resolved as for Act as me (a signed-in person on the roster, or the principal under local login; otherwise 403 `sign_in_required` / `not_on_roster`). `GET` returns `{notes, reply_notes, retention_days, effective_retention_days, company_retention_days, retention_choices, can_keep_notes, can_note_replies, can_set_company_retention, share_work_style, can_share_work_style}` (`can_keep_notes`: a team member on the People list may turn notes on; `can_note_replies`: they can use Act as me, so email replies are noted too; `share_work_style`: their own \"Share my work style with the team\" switch, `can_share_work_style`: a team member in a team workspace). `PUT` turns notes on only for someone who may keep them, and `share_work_style` on only for someone who may share it (403 `not_available` for either; turning either off always works), takes a person's retention only when shorter than the company's (422 `invalid_setting`), and the company retention only from the principal (403 `principal_only`). `PATCH` pins or corrects a note; `DELETE` forgets it; `forget-conversation` is \"Don't remember this\". A note that isn't the caller's is 404. Every change is audited privately to its person. |\n| `audit` | `GET /audit/logs`, `GET /audit/logs/{id}`, `GET /audit/sessions/{id}`, `GET /audit/usage`, `POST /audit/log` | Durable audit trail of side effects. `/audit/sessions/{id}` adds a per-session token cost summary; `/audit/usage` aggregates token usage and OpenRouter cost across all sessions (totals, by day, by model, and by source — the part of the system that made the calls — including server-side web-search counts), plus `scripts`: how many `run_script` scripts ran (and worked, and in workflows), the tool calls they made, their time, and `turns_avoided` — an upper bound on the model turns those calls would otherwise have needed (each call past a script's first) — and `python_jobs`: how many Python jobs ran (and worked, ran a kept Custom tool, took a chat attachment), their time, CPU, largest peak memory and bytes in and out, and the tokens and charge of the turns that ran one. The Token usage page shows it as a Scripts section. Rows private to the principal (`private_to_principal`: a turn about their private mail, or a row naming one of their contacts on their own turn) are theirs alone: for any other caller, resolved as for `/alerts`, the usage totals leave them out. Once a principal is on the roster, anyone else reads only their own rows (a teammate's own turns, their Act as me work) in the list, by id and in a session, and the principal reads those as actions only — what was done, when and in whose conversation, with the text, details and `full` left out, and no text search over them; any other row's id is a 404, and a session shows only the rows the caller reads. |\n| `architecture` | `GET /architecture/sections`, `.../sections/{id}` | This documentation surface. |\n| `guide` | `GET /guide/sections`, `.../sections/{id}` | Plain-language user-guide content surface — mirrors `architecture` (static, version-controlled files; no LLM on the serving path). |\n| `health` | `GET /health`, `GET /health/honcho` | Liveness probes. Only `/health` is unauthenticated; `/health/honcho` sits behind the gate like everything else, because the exemption is an exact-path set. `/health` also carries `google_tools_missing`, the pinned Google tools startup priming could not find: `[]` once it has checked and they are all there, a list of names when a workspace-mcp rename or the configured services and tool tier left one out, and `null` for no answer — Google Workspace is not configured, or the probe has not finished or it failed. Only `[]` asserts they are present. It also carries `gateway`, the MCP gateway's state from its supervisor (`orchestrator.gateway_boot`): `state` is `off` (MCP not enabled), `no_servers` (enabled, the config names none), `starting` (an attempt is running), `up` or `down` (every attempt so far failed; `next_retry_at` is the next one), with `attempts`, `up_since` and `email_poller` (whether the poller that rides on the gateway runs); the failure's own text stays out, since the route is open to anyone, and goes to the log and the owner's notice on Home. A hosted instance whose gateway never came up used to answer `ok` here with no email going out or in. Only `[]` asserts the tools are present, so a probe that died cannot read as healthy; Setup status says which of the three `null` means. |\n| `version` | `GET /version` | The running version, the latest release on GitHub and `update_available`, for Settings → About. The GitHub answer is cached 6h, only a semver `tag_name` is used and the release link is built from it; `UPDATE_CHECK_ENABLED=false` skips the call and still returns the running version. No in-app updater: upgrading is host-side (`docs/deployment.md#upgrading`). |\n| `setup_status` | `GET /setup/status` | Settings → Setup status: one light per part of the install — AI model, company setup, owner, the Executive's address, API protection, Gmail, your own mailbox (Act as me), Slack, Discord, Telegram, Google Chat, the scheduler, the daily brief (when and where it goes) and Honcho — as `{checked_at, checks: [{id, label, state, summary, fix, link, last_activity}]}`, where `state` is `ok`, `warn`, `error` or `off` (optional and not set up), `fix` says what to do in plain language, `link` is an in-app page that helps (`/people`, `/onboard`) and `last_activity` is the channel's latest `integration_inbound` audit time. It tests live, with at most one free read-only call per service, each bounded to 5 s: Anthropic `models.list` (so an organisation key missing `ANTHROPIC_WORKSPACE_ID` shows up), Slack `auth.test`, Telegram `getMe` and `getWebhookInfo`, Discord `users/@me` only while the bot is still connecting, and — once the owner's own Gmail credential exists — Google's token refresh and Gmail `users/me/profile`. It also reads the listeners the lifespan leaves on `app.state` (Slack's socket, the Discord bot and its task, the MCP gateway), the scheduler heartbeat and the pause switch. When `google_workspace` is configured, the lifespan also starts `MCPGateway.prime_pinned_tools()` in the background (held on `app.state.mcp_prime_task`, cancelled at shutdown), so the pinned Google tools named in the prompt are callable without a search. With `DRIVE_SYNC_ENABLED` the lifespan also bootstraps the `drive_sync_scan` heartbeat beside Notion's (see Scheduler); the Drive folder sync has no HTTP route, only the `openexecutive sync-drive` / `purge-drive` CLI. With `CONFLUENCE_SYNC_ENABLED` it bootstraps the `confluence_sync_scan` heartbeat the same way; that sync is run by hand with `openexecutive sync-confluence` / `purge-confluence`. Sample values copied from `.env.example` are named as such without being sent anywhere. No token reaches a response or a log line: probes report a status code, a service's short error code or an exception's class name. Requests within 5 s of a run get that run's answer and concurrent requests share one run (kept per app on `app.state`), so the button can't become a stream of calls to those services. Gated like every other route; the UI adds a sign-in light of its own, from its own settings. |\n| `documents` | `POST\\|GET\\|DELETE /documents`, `GET /documents/{filename}`, `GET /documents/sources`, `POST /documents/sources/{drive\\|onedrive\\|notion}/sync`, `GET /documents/{drive\\|onedrive\\|notion}`, `GET /documents/{drive\\|onedrive\\|notion}/{id}` | Upload, list, view extracted text, and delete indexed company docs. `GET /documents` rows carry `source: \"upload\"` and the `domain` the upload was indexed under. The connected-source routes back the knowledge page's single Company documents list: `/documents/sources` gives each source's `{id, label, enabled, syncing, last_run, last_error, interval_minutes, file_count}`; `/documents/drive`, `/documents/onedrive` and `/documents/notion` list synced files from the sync state (`{enabled, last_run, last_error, files: [{id, name, url, modified_at, synced_at, indexed}]}`); `/documents/{drive\\|onedrive\\|notion}/{id}` returns the stored text (a OneDrive id is `<drive id>:<item id>`) for an id on record (the path comes from the state, never the caller). They are read-only — a synced file is removed where it lives. `POST …/sync` starts one tick of the same sync the scheduler runs as a background task (202); 409 when the source is not connected or a tick is already running (every sync holds a per-process run lock, so a manual run never overlaps the heartbeat; the route also keeps its own started task per source and re-checks just before starting one, so two simultaneous requests start one tick, not two), 429 with `Retry-After` within 60 s of the last successful run or of the end of the last tick in this process, failed ones included, so a failing or slow source can't be re-run back-to-back. The page's text view never loads images from synced content (a remote image would be a read beacon for whoever edits the shared folder). Upload is keyed on the filename, so re-uploading a document replaces its chunks rather than duplicating them, and `DELETE` removes the rows as well as the file. `domain` must be one of the nine specialist domains (including `sales`) or `general`; anything else is a 400 rather than a 200 that indexes where no specialist can retrieve. A scanned PDF is converted (`knowledge.pdf_reader`) before it is indexed and in the text view, as are scanned PDFs in `POST /chat/upload` and the client / onboarding intake uploads. The upload is streamed to disk 1 MB at a time and refused with a 413 once it passes 50 MB, and the file is parsed in a short-lived child process (see the RAG section), so a large document does not stay in the API process's memory. A Word or Excel file that cannot be parsed, or whose parse runs out of time, is a 422 with a readable reason. When every parser is busy, the response is a 503 with `Retry-After`. |\n| `executive` | `GET /executive/status`, `POST /executive/pause`, `POST /executive/resume` | Operator pause switch for all autonomous work (scheduler, Gmail poller, workflow resumer); inbound chat and DMs stay live. Status returns `{paused, paused_at, paused_by, reason, held_actions, can_resume}` — `held_actions` counts due `pending` rows that fire on resume. `pause` takes an optional `reason` (≤200 chars) and is idempotent (re-pausing keeps the original start and reason); `paused_by` is the UI proxy's `x-caller-email`, else `api`. Behind the shared-secret middleware — not `SCHEDULED_ADMIN_TOKEN`, which the UI cannot send. Any caller may pause; `resume` is principal-only (`is_principal_or_self`, 403 otherwise; open while no principal is rostered; a header-less caller resolves to the principal as in `chat`), and status carries `can_resume` for the viewer so the UI hides Resume from everyone else. |\n| `workspace` | `GET /workspace`, `PUT /workspace` | Solo/team mode, the user's time zone and the principal's role (`memory.workspace_settings`). Both return `{mode, timezone, effective_timezone, role_kind, role_title, reports_to, remit, measured_on, company_domains, company_domains_custom}` — `timezone` is what the user set (null when unset), `effective_timezone` the zone in effect (`timezone`, else `USER_TIMEZONE`, else UTC), and each role field is null until set. `PUT` is a partial update: only the fields sent change; `mode` is `solo` or `team` (not null), `timezone` an IANA zone, or null/blank to clear it; `role_kind` is `owner`, `in_house`, `independent` or `other`, and `role_title` / `reports_to` / `remit` / `measured_on` are free text (trimmed, at most 120 / 120 / 500 / 300 characters), each null or blank to clear it; an unknown zone, a bad mode or role kind, an over-long role field or an unknown field is a 422 and changes nothing. Switching to solo cancels every pending department check-in, switching to team re-schedules them, and a new zone re-times the pending briefs and reflection; a role change has no side effects (the next turn reads it). `company_domains` is the company's own email domains (a teammate matches by local part on them — see Org): `GET` returns the ones in effect (the stored list, or the domain of the principal's primary address, with `company_domains_custom` false) only to the same caller the role is shown to, `[]` for anyone else; `PUT` takes a list (at most 10 plain domain names, never a free-mail provider's — 422 otherwise, the value never quoted) or null / `[]` to go back to deriving them, and a change is audited with the new list. Each change writes a `workspace_settings_changed` audit row (a role change adds `details.role_fields_changed`, the names of the fields that changed — never their text). The role is the principal's: `GET` returns it only to a caller `PUT` would accept and reads it as nulls for anyone else. `PUT` is the principal's alone (`is_principal_or_self`, 403 otherwise; a header-less caller resolves to the principal as in `chat`) and open while no principal is rostered, so first-run setup can pick a mode. Both work before onboarding — unlike `company-profile`, nothing here needs a profile. |\n| `scheduled` | `GET\\|DELETE` scheduled actions (`?status=…&order=asc\\|desc`) | Pending and fired scheduled actions; `order` flips the `run_at` sort (asc=soonest-first for the upcoming queue, desc=most-recent-first for terminal-status history). `DELETE /scheduled/{id}` cancels a `pending` row (409 otherwise). Like pause it only stops work, so any signed-in user may — the UI proxy's `x-caller-email` is trusted when the request also carries the `BACKEND_SHARED_SECRET` `x-api-key` — as may a matching `X-Admin-Token` (`SCHEDULED_ADMIN_TOKEN`) or a loopback caller with no token set; otherwise 401 (token set) or 503. Audited as `scheduled_action_cancelled`. |\n| `watchlist` | `GET\\|POST\\|PATCH\\|DELETE /watchlist`, `POST /watchlist/{slug}/approve`, `POST /watchlist/{slug}/decline` | External-monitoring watchlist + signals. Rows carry `origin` (`manual` \\| `executive` \\| `research` \\| `research_proposed`); `approve` turns a research suggestion (`research_proposed`, `dry_run`) into a live watch, `decline` `{reason: not_relevant \\| too_noisy \\| wrong_source}` removes it and records the decline (or, for `too_noisy`, keeps it live at a high floor), both 409 unless the row is a pending suggestion; `DELETE ?reason=` on a research-origin row records the decline too. |\n| `workflows` | `GET /workflows`, `POST /workflows/{name}/runs`, `GET\\|POST\\|PUT\\|DELETE /workflows/custom`, `GET\\|DELETE /workflows/custom/{name}/targets`, `POST /workflows/runs/{run_id}/decision`, `GET /workflows/tools/search?q=`, `GET /workflows/tools/describe?names=` | Workflow catalog, run management, and user-created (dynamic) workflow CRUD. Custom definitions come back with a server-managed `owner_person_id` (the creator — the web caller on `POST`, set on insert only; a client-sent value is ignored), who is asked to approve a workflow's first write to a new target. `GET /workflows/custom/{name}/targets` lists the targets its tool steps may write to without asking and `DELETE …/targets?value=` forgets one. `POST /workflows/{name}/runs` is a 403 (\"Only the principal can run this workflow.\"), audited, before any run row exists, when the workflow is one only the principal may start in the workspace's mode (`weekly_review` always, `morning_brief` in solo) and the caller is not the principal. A request with no `x-caller-email` counts as the principal; with no principal on the roster, no signed-in caller does. This governs starting a run only. A run started by hand (this endpoint, or `run_workflow` / research in chat) records its starter as `owner_person_id` and is theirs alone: `GET /workflows/runs` lists the team's (scheduled, no owner) runs plus the caller's own and the ones waiting on them, and `GET` / `DELETE /workflows/runs/{run_id}` and `POST /workflows/runs/{run_id}/decision` are a 404 for anyone else's, the principal included. A signed-in email that is on nobody's People entry cannot start one (403), and a run whose starter can't be read is refused (503); its steps read and draft documents as the starter only. `POST /workflows/runs/{run_id}/decision` with `{decision: \"approve\"|\"reject\"}` answers a run awaiting a yes/no sign-off from the web app — 403 unless the caller is the person it waits on or the principal, 409 if it isn't awaiting or needs a written reply — through the same `apply_resolution` a chat reply uses. Custom create/update also rejects (422) any action-step tool that does not resolve, and so does `POST /workflows/custom/{name}/activate` when turning a workflow **on** — that click is the approval for a tool workflow chat saved switched off (turning off is never checked). Turning on a workflow with action steps requires the body `{is_active: true, definition}` carrying the definition the user reviewed; if the stored one differs (ignoring `is_active` and timestamps) it returns 409, and the switch itself is a compare-and-set on the validated revision (`dynamic_store.activate_if_unchanged`, keyed on `updated_at`), so a save that lands while the tool check awaits the gateway — even from another process — is a 409 too, so only what was shown gets switched on. `GET /workflows/custom` lists inactive definitions too; `GET /workflows/{name}` 404s for them. The two `tools` endpoints return `{tools: [{name, description, read_only, source}]}` for the advanced editor's tool picker and the review card; `read_only` is true only when the name says it reads (a label, not a control). The run stream adds a `progress` event (`step_id`, `summary`) while an action step uses tools. Each catalog entry (`WorkflowMeta`) carries `is_custom`, `background` (true for the system-run built-ins the UI files under “System”) and `quick` (true for a Quick workflow, one per playbook nothing else follows, named `quick_<playbook>`). The run endpoint streams SSE; its `awaiting_human` event carries `delivery` (`self` \\| `sent` \\| `suppressed` \\| `alerted` \\| `failed`) alongside `person_id` / `question` / `awaiting_until`, because the outbound guard or an unreachable approver means the question may never have been asked — a client must not render “waiting on them” for those. A paused run ends the stream with neither `done` nor `error`, so that `awaiting_human` frame carries `terminal: true` to say so explicitly, plus `resumable` — whether the run continues by itself once answered or stops at the gate. `GET /workflows/runs/{run_id}` returns the run row with the resume payload replaced by a small `resume_progress` summary (`gate_step_id`, `completed_step_ids`, or `null`): the payload holds every completed step's full text, and the run-detail page polls this endpoint while a run is unfinished. |\n| `knowledge` | `/knowledge/builtin`, `/external`, `/search` | MBA knowledge base and external sources. `/search` is the introspection parallel of `retriever.retrieve` and mirrors its filters — including the `general` catch-all on the company collection — so the panel cannot report a specialist seeing nothing that it does retrieve in chat. It accepts `general` in `domain_filter` for the same reason. `/search` applies the same review gate as `retrieve()` — withheld (pending or rejected) built-in, external and failure rows are dropped, so the panel never advertises knowledge the Executive cannot use; it degrades to unfiltered rather than failing if review state is unreadable. `POST /knowledge/builtin` refuses a name already held by shipped content, and `POST|DELETE /knowledge/failures` now register and deregister for review — without that a user-authored failure case study had no review row, so no SME decision could block it. |\n| `skills` | `/skills`, `/skills/search`, `/skills/{name}/restore` | Skill registry (the UI's **Playbooks**) CRUD and search. `PUT` on a built-in saves a company customization that shadows it; `DELETE` returns `{name, outcome}` — `deleted`, `reverted` (customization removed, built-in back) or `hidden` (built-in hidden for this company); `restore` un-hides; `GET /skills?include_hidden=true` lists hidden built-ins. List and detail carry `used_by` — `[{name, title, is_custom}]` of the workflows that follow the playbook (switched-off custom ones included) (`WorkflowMeta.playbooks` is the other direction). |\n| `evals` | `/evals/scenarios`, `/evals/runs` | Eval scenarios and run results. `POST /evals/runs` is a 403 when a workflow scenario names a workflow that only the principal may start in the mode it runs in (the weekly review; the morning brief in solo) and the caller is not the principal — the Jobs page's rule for starting a run; stored results are read under the existing `/evals/runs` rules. |\n| `memories` | `/memories/{decisions,initiatives,advice}`, `GET /memories/facts`, `POST /memories/facts/{fact_id}/{retire,approve,decline}`, `GET /memories/facts/approval`, `PUT /memories/facts/approval/{person_id}`, `GET /memories/people`, `GET /memories/people/{person_id}/conclusions` | Episodic-memory record management; the standing facts the Pulse Corrections tab reads (`{facts: [...], can_retire, retirable_ids, can_review}`; each fact carries `recorded_by_role` / `recorded_by_name`; `?include_inactive=` default true, honoured for the principal only; anyone else gets the active rows plus their own `proposed` ones, with the provenance (`session_id`, `turn_id`, `recorded_by_person_id`, `retired_reason`, and `source_quote` except on their own facts) blanked, never retired, replaced or declined text; `?limit=` 1–500) and their retire (the principal any fact, a teammate only one they recorded that the principal has not approved — else 403; the audit row names the teammate; 404 for an unknown or `profile` row, 409 when no longer active; writes a private `fact_retired` audit row); approve / decline a teammate's `proposed` fact (principal only, else 403; 409 when it is not proposed; private `fact_reviewed` row); and the per-teammate \"needs my approval\" switch (on by default; `GET` lists `[{person_id, full_name, needs_approval}]` for every active team member but the principal, `true` for anyone not yet set; `PUT` takes `{needs_approval}`, 404 for anyone else; both principal only; private `fact_approval_changed` row); plus the read-only peer-memory overview the Pulse People tab reads (`{status: ok|disabled|error, people: [...], conclusion_total}`; `?recent=` 1–50, default 5) and one person's full conclusion list, paged newest first (`{status, person_id, items, page, size, total, has_more}`; `?page=` ≥1, `?size=` 1–100, default 50; 404 when the person has no peer). Peer memory is each person's own: the overview holds only the caller's entry (none for a caller off the roster; `conclusion_total` counts only theirs), the principal included, and `/conclusions` for anyone but the caller is a 404. |\n| `review` | `/review/items`, `/review/stats`, `/review/trusted-defaults`, `POST /review/curate`, `POST /review/bulk-approve` | Knowledge review queue and annotations. Shipped knowledge registers as a trusted default (`approved` with a NULL `reviewed_at`), so the queue starts empty; `curate` (`{domain, action: start|stop}`, domain required) opts one domain in or out, and `trusted-defaults` returns domain → unreviewed count. `bulk-approve` takes a non-empty `domain`, `item_ids` (max 500) or `all_pending: true`, and 400s with no selector; it returns `{approved_count, item_ids}`. |\n| `onboarding` | `/onboard/interview/understand`, `/onboard/interview/start`, `/onboard/interview/message`, `/onboard/interview/draft`, `/onboard/interview/{id}`, `/onboard/interview/commit` | Conversational company setup. The `onboarding_interviewer` agent asks clarifying questions, then emits a draft profile + people + departments. **Nothing is persisted until `POST /onboard/interview/commit`** — the interview itself never writes. `start` is multipart (description + optional deck/one-pager uploads). `commit` answers 403 when the draft would make a different person the owner and the caller is not the current owner (setup demotes the old principal, and any signed-in user can open it). It also takes an optional `owner_email` — the principal's sign-in email, which the review screen pre-fills from the signed-in user — checked before anything is written (422 if it is malformed, already another person's, or would replace the principal's existing email — setup only fills in a missing one; 403 if the caller is not the owner and it is not the address they signed in with, since the roster is the sign-in allow-list) and saved on the principal's People row, so the owner's chats resolve to them from the first turn. |\n| `workflow_designer` | `POST /workflows/designer/start`, `POST /workflows/designer/edit`, `POST /workflows/designer/message`, `POST /workflows/designer/draft`, `GET /workflows/designer/{session_id}` | The conversational New workflow wizard, and editing a saved workflow by conversation. `/edit` (`{name}`) opens a session on a saved custom workflow with no model call: the stored definition is the starting draft, and every response then carries `editing` (its name) and `original` (the stored definition as the session opened on it); drafts keep that name and its on/off state whatever the model writes (404 for an unknown workflow). The UI saves one with `POST /workflows/custom/{name}/save-edit` (`{session_id, definition}`): the server saves the session's own draft, never a client-sent body, and only when it equals the `definition` the user reviewed (a later message is a 409, so the click saves only what was shown); it re-validates it, keeps the stored on/off state, and writes only if the stored workflow is still the version the conversation opened on — otherwise 409, as it is for a session about another workflow or one with nothing drafted yet. It sits under `/workflows/custom` so anything that limits writes to custom workflows covers it too. The `workflow_designer` agent asks clarifying questions (`phase: \"question\"`, with `hint` and up to four suggested `options`), then returns `phase: \"draft\"` with `draft.definition` (exactly the body `POST /workflows/custom` takes), `summary`, and `assumptions`; every response carries the `transcript`. **Nothing is persisted here** — sessions are in memory, and the UI saves through `POST /workflows/custom`. Errors are fixed strings (422 over the 8,000-char message cap or the transcript cap, 404 for an expired session, 409 while another turn on the same session is still running, 502/504 for model failure or timeout). Mounted ahead of the `workflows` router. |\n| `python_jobs` | `GET /python-jobs/{job_id}/{name}` | A result file a Python job (`run_python_job`) wrote, the principal's alone (403 otherwise), served as a download (`Content-Disposition: attachment`, `Content-Security-Policy: sandbox`, `nosniff`, `Cache-Control: no-store`); 404 for a bad id or name or a file past its 7 days. |\n| `saved_tools` | `GET /saved-tools`, `GET\\|PUT /saved-tools/{name}`, `POST /saved-tools/{name}/rollback` | The scripts the Executive saved as tools (Settings → Advanced → Saved tools), the owner's alone (403 otherwise). The list gives each tool's name, description, on/off, current version, `kind` (`script` or `python`, a kept Python job) and the tools it calls, plus whether `SAVED_TOOLS_ENABLED` is on; the detail adds the script, every version and the last 20 runs. `PUT {enabled?, workflows?, version?}` turns one on or off (off also turns it off for workflows, so turning it back on never re-arms unattended runs), and turns the named `version` — the one the owner looked at, never whatever is current by then — on for workflows (or workflows off); a Python tool never turns on for workflows (409) — turning workflows on lets more run unattended, so it needs a request tied to the owner (signed sign-ins or local login, as Take the lead's switches do; 409 otherwise); `rollback {version}` makes that version the one that runs. Both are audited as `saved_tool_changed`. The Executive saves tools itself through `run_script`; there is no create route. |\n| `onboarding` (fallback) | `/onboard/start`, `/onboard/answer`, `/onboard/status/{id}` | The original 12-step form wizard, kept as a fallback (UI at `/onboard?mode=form`) and used by the `openexecutive onboard` CLI. Its final answer saves the profile and adds people with their emails and a principal, so once there is an owner only they can send it (403, and the session rolls back so the owner can still finish it). `/onboard/start` reads the workspace mode once for the session: solo skips the Team Size, Team Members and Fractional Executives steps. Each status response's `current_step` / `total_steps` count only the steps that session asks, and `optional` says whether the current step can be skipped. |\n| `company-profile` | `GET\\|PATCH /company-profile` | Structured company profile. |\n| `alerts` (private cards) | `POST /alerts/{id}/ack`, `/reopen`, `/alerts/bulk-ack` | An alert tagged `private:principal` (see Today) is a 404 on ack and reopen for anyone but the principal, and their bulk sweep skips it. A roster-request card (source `roster_request`) is a 409 on `ack` even for the principal — it is answered at `/people/requests/{id}`, since acking would leave the sender waiting with nothing to show it — and, as an exempt source, `bulk-ack` and `reopen` skip it. |\n| `auth` | `GET /auth/allowed-emails` | Non-archived **team members** with an email, offered to the UI's sign-in gate — a contact is never on it, so adding a client to reach them by email no longer lets them sign in. The gate takes the **union** of this roster and its own `ALLOWED_EMAILS` env list, so env entries always work (and are checked first, short-circuiting the fetch) and roster entries are additive — a roster change can never revoke a configured operator. With local login (`make dev` with Google sign-in not set up — the UI then listens on 127.0.0.1 only) the UI signs in without an email, sends no `x-caller-email`, and the backend resolves the caller to the principal, as it does for the CLI. |\n| `fixtures` | `/fixtures`, `/fixtures/{name}/load` | Demo-company fixture load/reset/snapshot. `POST /fixtures/reset` leaves no departments, like a new install (`departments_seeded` is always 0). |\n| `clients` | `/clients`, `/clients/generate`, `/clients/cockpit`, `/clients/save`, `/clients/{slug}/activate` | Client-company slots — multi-client mode for fractional use; create from current/blank/AI-generated intake draft (pasted notes and/or file attachments, multipart on `/clients/generate`), save-back switching, engagement metadata (PATCH), cross-client cockpit, delete. See the Client Companies section. |\n| webhooks | `POST /webhook/telegram`, `POST /webhook/google-chat`, `POST /webhook/slack` | Inbound integration callbacks (self-verified). `/webhook/slack` answers 503 unless `SLACK_SIGNING_SECRET` is set. |\n| MCP server | `POST\\|GET /mcp` | Open Executive exposed as an MCP server (Streamable-HTTP). Not a router — a mounted FastMCP sub-app. Same shared-secret gate (`x-api-key`); exposes company context as resources and the specialist council as tools. See the MCP Server section. |\n\n**Take the lead** (`api/routes/take_the_lead.py`): `GET /take-the-lead` returns the As the Executive switch, its six Ask first switches (`{kind, label, hint, on}`, the hint a plain line on what each one catches), the company's rules, whether this server can tie a change to the owner (`available`), and whether the Executive is paused. `PUT /take-the-lead` takes `{enabled?, ask_first?}`, and `POST /take-the-lead/rules` / `DELETE /take-the-lead/rules/{id}` add or remove a company rule (`{kind: person|domain|words|amount, value}`, 422 on a bad one). All four are the principal's alone (403). Anything that lets more happen on its own (turning it on, turning an Ask first off, removing a rule) needs a caller the API can tie to them (`delegation.verified.caller_refusal`, else 409). `GET /today/activity` adds `took_the_lead` rows (what Take the lead did, named, not quoted) and, for the caller alone, `sent_as_you` rows for what Handle it for me sent as them. Under `/delegation`: `PUT /delegation/take-the-lead {enabled}` is Take the lead as you (the owner's only for now; turning it on turns Handle it for me on), and `GET|POST /delegation/take-the-lead/rules` and `DELETE …/rules/{id}` manage the caller's own rules. `HandleItOut` gains `lead` and `lead_available`. A `take_the_lead_action` decision is approved or rejected through the existing `/decisions/{id}/approve|reject`. Training: the response also carries `training` (In training: everything waits unless allowed), `learned` (`{id, label, feature, example, uses, created_at}`: what it was allowed, the edit kept as an example by field) and `suggested` (`{key, label, approvals}`); `PUT` takes `training?` too, and leaving training while on needs the same provable owner as turning it on. `POST /take-the-lead/learned {key}` allows a current suggestion (404 for anything else; provable owner), `DELETE /take-the-lead/learned/{id}` makes it ask first again. `learned` also lists the owner's own Act as me senders (`feature` `act_as_me`), never anyone else's, and Remove works on those too. On a `take_the_lead_action` card, `POST /decisions/{id}/approve` takes `edits.input` (the card's editable fields, 422 for any other; the principal's alone, 403 for an area approver) and `edits.allow: true` (a training card only, 422 otherwise; the principal from a provable request, 403 otherwise), finishing `approved_with_edit` when a field changed.",
"mermaid": null,
- "generated_at": "2026-10-13T11:57:15Z"
+ "generated_at": "2026-10-13T15:30:00Z"
}
diff --git a/packages/core/openexecutive/architecture/prebuilt/audit.json b/packages/core/openexecutive/architecture/prebuilt/audit.json
index eac0f617..d371e5a6 100644
--- a/packages/core/openexecutive/architecture/prebuilt/audit.json
+++ b/packages/core/openexecutive/architecture/prebuilt/audit.json
@@ -1,7 +1,7 @@
{
"section_id": "audit",
"title": "Audit Log",
- "markdown": "The audit log is a persistent, searchable, append-only record of what the system did — distinct from episodic memory (which biases future LLM behavior). Every event flows through `audit/logger.log_event`, a synchronous SQLite writer (WAL + 5s busy_timeout) that appends one row to the `audit_log` table in `episodic_memory.db`.\n\n### Event types\nRows are tagged by `event_type`, including `chat_turn`, `specialist_consult`, `tool_invocation`, `scheduled_action`, `alert`, `integration_inbound` (Slack / Discord / email / Telegram / Google Chat), plus `knowledge_retrieval`, `fact_confirmation` (an emailed standing-fact change held, applied, cancelled or refused; private), `fact_reviewed` (the principal approved or declined a teammate's proposed standing fact; private), `fact_approval_changed` (the principal turned \"needs my approval\" on or off for a teammate; private), `fact_retired` (the principal, or the teammate who recorded it, retired a standing fact from the Pulse page; private, since it names the fact a teammate can no longer read; the fact tools' own rows and their `skill:` dispatch rows are private `tool_invocation` rows, since they quote the principal and name their session and turn), `memory_extraction` (one row per episodic-extraction pass, written in a `finally` so a crashed pass is recorded too: how many items the model proposed, how many were stored, a `{kind, reason}` record per drop, a `malformed` count for payload shapes that were never usable items, and a `failure` naming the exception type when the pass raised — including `CancelledError`, which is a `BaseException` and so would otherwise be logged as a clean empty pass — a `proposed>0, stored=0` streak means the quote validator is rejecting everything, which is otherwise invisible), `cache_event` (one row per model call from every call site — Executive turns, specialist consults (chat turns and workflow steps), research specialists, the research routing and watchlist passes, triage, memory extraction, alert review, the brief's narrative and its grounding repair, the digest, reflection and weekly-review workflows, committee reviewers, and the small utility calls (session titles, the chat-channel response gate, reply parsing, PDF transcription); a unit test fails any model call in the `openexecutive` package that records no row — carrying tokens, cache hits, cost, server-side search count and, inside a research run, its `run_id`), `memory_snapshot`, `committee_review`, `peer_memory`, `attunement` (one row per open-loop extraction pass — opened / closed / dropped with a reason per drop and a `failure` type — plus a row per loop opened or closed and per 👍/👎 on a reply; see Attunement), `executive_paused` / `executive_resumed` (the operator pause switch; actor is the caller's email, details carry the reason, or on resume the held-action count), `scheduled_action_cancelled` (a pending scheduled action cancelled from the Pulse page or a script; actor is the signed-in email, `admin-token` or `local`), `workspace_settings_changed` (solo/team mode, the user's time zone or the principal's role changed through `PUT /workspace`; actor is the caller's email or `api`, details carry the mode's and zone's from/to, plus `role_fields_changed` — the names of the role fields that changed, never their text — when the role changed, and the new `company_domains` list when the company's email domains changed), `decision_class_mode_changed` (a decision class switched between `propose` and `auto_execute` through `PUT /decisions/classes/meeting_scheduling`; same actor rule, details carry the class and the mode's from/to), the Act as me events — `delegation_settings_changed` (a person turned it on or off for themselves), `delegation_reminder_set` and `delegation_reminder_sent` (a person set a reminder for themselves with `remind_me`, and it went to them; ids and times only, never the text; private to that person), `delegation_actions_proposed`, `delegation_actions_approved` and `delegation_actions_dismissed` (the Executive left a person an approval card, which they approved or dismissed; kinds and outcomes only, never the text; private to that person), `delegation_gmail_verified` (their own Gmail checked before turning it on), `delegation_voice_changed` (\"How I write\" learned, edited, locked or reset, or its signature read from Gmail again), `delegation_voice_described` (a style written from the person's own description, counts only; nothing saved until they save it as an edit), `delegation_drafted` (one per draft saved in their Gmail from chat), `delegation_inbox_changed` (Draft replies to my inbox turned on or off), `delegation_inbox_scanned` (one per inbox scan, with its counts and status), `delegation_reply_drafted` (the inbox watcher saved a reply and made its card), `delegation_reply_dismissed` (a card dismissed, and whether its draft was deleted or kept), `delegation_reply_closed` (a card Gmail already settled: sent or deleted there, answered, gone or expired) and `delegation_reply_sent` (the person tapped Send and the draft went, or the reconciler confirmed an unconfirmed send did) — always private and metadata only: ids, counts and flags, never the draft or the thread; the roster-request events (see Org) — `roster_request_created`, `roster_ack_sent`, `roster_ack_withheld` (an acknowledgement the adapter declined — an email Gmail did not authenticate), `roster_request_resolved` (actor is where it was answered: `web`, a chat channel, `email` or `people_page`), `roster_message_replayed` (one per held message: replayed, dropped, failed or unavailable) and `roster_email_answer_refused` (an emailed answer with no live token, or failing DMARC) — always private to the principal and never carrying what the new sender wrote, like the `integration_outbound` row (`kind: roster_ack`) for the one gateway send allowed to an address off the roster, the fixed acknowledgement; `grounding` (unattended prose that named people or figures its inputs don't hold — a brief or digest that held back lines or used its repair call, a reflection tool call refused or flag held back, a triage or review rewrite replaced; details carry the surface, the mode, the ungrounded names and figures and up to five held-back lines cut to 200 chars, private when the brief read the principal's private rows or the alert is private; in `GROUNDING_CHECKS=report` mode these rows are written and nothing is changed); and the external-signal/monitor events. An unknown chat sender's `integration_inbound` row still reads `rejected_unknown_sender` — it means the message did not reach the Executive, not that it was discarded: a DM or mention is held as a roster request.\n\n### Session & turn context\nEach row carries `session_id`, `turn_id`, `actor`, `summary`, `details_json`, `full_json`, `department`, `private_to_principal` and `private_to_person`. A `turn_id` (`f\"t-{uuid12}\"`) is set per Executive turn and propagated automatically via ContextVars to retriever / specialist / tool / cache emits, grouping one user-assistant exchange. `summary` is the searchable column (LIKE, ~300 char cap); `details_json` is structured and not searched. `department` is filterable in SQL (`AuditLogger.query(department=…)`) — the department check-in's skip rule uses it to ask whether anything was logged for a department since its Goals were last reviewed.\n\n### Rows private to the principal\nThe principal reads `/audit`, and others their own rows (see below), so a row about something private to the principal is marked `private_to_principal` as it is written, and only the principal reads it. `AuditLogger.log` marks a row private when its caller passes `private=True` — the email poller's two inbound rows for mail from one of the principal's contacts or mail they forwarded, and `create_alert`'s row for a private alert — or when `people_tools.audit_row_private_to_principal` says so at write time:\n- every row a turn private to the principal writes (`Session.private_to_principal`): its reply, each tool call, the model-call rows, and the background passes that inherit the turn;\n- on the principal's own verified turn (or inside `grant_contact_egress`), any row whose summary, details or full payload names one of their contacts — their email address or full name anywhere in the text, or their person id or chat id in a person or recipient field (`person_id`, `attendee_person_ids`, `user_id`, `chat_id`, `channel_ref`), JSON tool results included.\n- every row an Act as me turn writes once `ghostwrite_email` or a mailbox read (`search_my_email`, `read_my_email`, `read_my_email_attachment`, `my_email_awaiting_reply`) has touched the speaker's mailbox (`TurnDelegation.touched_mail`, set before the tool's first read): the tool row, the reply's `chat_turn` row carrying the draft's preview, the model-call rows, and anything it starts in the background, under `private_rows` (it runs no learning passes). Only the principal can have Act as me for now, so private to the principal is the right visibility; the tool's input and result are also redacted (`ghostwrite` and `my_email` are sensitive tool names).\n- every row the Act as me inbox watcher, a reply card's Send or Dismiss, an action card's Approve (or the card carried out on its own, in training), or a person's Act as me settings write: they run under `rows_for_person(person_id)` (or pass `private_to_person`), so their model-call rows and events are private, and the reply cards are their own person's alone. That includes what training records: `delegation_training_changed`, `delegation_reply_allowed`, `delegation_example_kept`, `delegation_actions_allowed`, `delegation_actions_handled`, `delegation_actions_held` and `delegation_learned_removed`.\n\n### Rows that belong to one team member\nAct as me can be on for team members as well as the principal (see Act as me). What a team member's own mailbox touches must not reach the principal either, so such a row also carries `private_to_person` — the team member's id — as it is written. A teammate's own conversation with the Executive is theirs the same way, as their mail is: every row of a turn a team member other than the principal speaks on a verified surface (the web app signed in, their own Slack or Discord, a secret-verified private Telegram chat) or by email Gmail authenticated as theirs (`people_tools.teammate_turn_owner`), its reply, tool calls, model calls and memory snapshot included. The rows written before the turn binds its session follow the same rule: the chat adapters' `audit_rows_on_senders_turn` holds `rows_for_person` for a verified teammate (`sent_by_teammate`), the web chat route for a signed-in teammate, and the email poller for authenticated teammate mail, which it checks before writing its inbound row. An unproven sender (a forged From, an unverified Telegram chat) keeps the ordinary rule, so nobody can hide a turn from the principal by claiming to be a teammate. `people_tools.audit_row_owner` decides it: the id a caller passes, else the `rows_for_person` scope, else the speaker of a turn whose `touched_mail` is set, else the teammate whose own turn it is; the principal's own id is never stored, so their rows stay private to the principal exactly as above, and a principal lookup that fails keeps the owner. A row with an owner is always private, and `AuditLogger.query` and `count` leave it out in SQL unless the caller asks for that person's rows (`owned_by`; `only_owned` keeps those alone, as the read routes do for a teammate) — so the brief's handled-overnight block and every other internal reader skip it. The read routes show it to the principal as an action only (`api.routes.audit._redact`): the summary is rebuilt from the event type and a plain tool name with whose conversation it was (\"mcp:send_gmail_message (Ben's conversation, details private)\"), the details keep only what was done and what it cost (`tool`, `kind`, `ok`, `refused`, token counts, …), and `full` is dropped; a text search (`q`) leaves such rows out, so their words never decide a match. The person it belongs to reads it in full. Only `usage_summary` counts it, so spend stays whole. `usage_summary` also adds up the sandboxed scripts (`run_script`) from their own rows — chat's `tool_invocation` with `kind: script` and a workflow step's `workflow_tool_call` with `tool: run_script`, each carrying `calls` and `duration_ms` — into `scripts` (scripts run and worked, tool calls, time, and `turns_avoided`, an upper bound on the model turns those calls would otherwise have needed); a script's individual calls are not counted twice. It also sums the `python_job` rows (one per `run_python_job`: `ok`, `duration_ms`, `cpu_ms`, `peak_mb`, files and bytes in and out, `attachments`, `saved_tool`) into `python_jobs`, with the `cache_event` usage of the turns that ran a job, each turn counted once. Rows that quote someone's own documents follow the same `rows_for_person` scope: a `draft_artifact` / `get_artifact` / `list_artifacts` dispatch row, the document tools' own rows, and a turn's `memory_snapshot` when its retrieval recalled a published document are private to the speaker (to the principal when they are the speaker).\n\nSome rows are written before their turn binds its session, so two scopes in `audit.context` count too. The email poller holds `private_rows` for the whole handling of a mail from a contact or a principal forward: every row is private, the knowledge retrieval (which quotes the mail) included. The chat adapters hold `principal_turn_rows` once they know the principal sent the message on a verified surface (`people_tools.audit_rows_on_senders_turn` on Slack, Discord and Telegram; the web chat route checks its caller), so the inbound row, the knowledge retrieval, alert triage and the start of the web stream follow the principal-turn rule above. Tasks started inside inherit the scope, except unattended work: a workflow run a chat reply resumes and each scheduled action start with neither (`unscoped_audit_rows`), so their rows follow the ordinary rule. A sender lookup that fails counts as not the principal, as the turn's own check does. Neither scope lets the turn reach anyone. A caller can add privacy but never remove it, and a check that cannot run keeps the row private. On anyone else's turn a contact is a stranger, so a row that names one keeps its visibility: hiding it would tell whoever wrote it that the address belongs to a contact. The poller handles each message with no session bound, so one message's private turn never marks the next message's rows.\n\nEvery read route resolves the caller as `/alerts` and `/decisions` do (`people.caller_is_principal`: the principal's signed-in email, or no `x-caller-email` — the CLI, direct curl, local login — once a principal is on the roster; before anyone is principal, nobody), in `api.routes.audit._view`. The principal reads every row, a teammate's own as an action only (above). Once a principal is on the roster, anyone else reads their own rows (`owned_by` = them, `only_owned`) and nothing more, since the log holds what everyone said to the Executive; a caller the roster can't resolve reads none. Before anyone is principal, everyone reads the rows not private to one. All of it is filtered in SQL, so `total` and pagination stay honest: `GET /audit/logs` leaves the rest out, `GET /audit/logs/{id}` answers 404 exactly as for a missing id, and `GET /audit/sessions/{id}` shows only the rows the caller reads (404 when none are left). `GET /audit/usage` is spend, not content: for anyone but the principal it leaves out the rows private to the principal, as before. Internal readers see every row unless they ask otherwise; the department check-in, whose report goes to the department, reads its recent activity without them, and its \"nothing new since the review\" skip rule ignores them too. Ids are one autoincrement sequence, so anyone else can still see a gap where a private row was written; that is accepted.\n\nAlways in the loop writes `history_notes_added` (count and source, never the note) and `history_settings_changed`, `history_note_changed`, `history_note_forgotten` and `history_conversation_forgotten` from its routes. Every one is private and belongs to its person alone (`private_to_person`). `recall_history` is a sensitive tool name, so its input and result are redacted.\n\n### Views\n- `GET /audit` — filterable list with a \"Group by session\" toggle and session pills.\n- `GET /audit/session/{id}` — full ordered timeline plus a server-derived node/edge graph rendered as a three-pane React Flow chart.\n\n### Invariant\nAudit writes never break a chat turn: all writes swallow exceptions and log a warning, and instrumentation observes only — it never touches request building or the cached prompt blocks. Audit is append-only: no UPDATE or DELETE paths exist in product code.",
+ "markdown": "The audit log is a persistent, searchable, append-only record of what the system did — distinct from episodic memory (which biases future LLM behavior). Every event flows through `audit/logger.log_event`, a synchronous SQLite writer (WAL + 5s busy_timeout) that appends one row to the `audit_log` table in `episodic_memory.db`.\n\n### Event types\nRows are tagged by `event_type`, including `chat_turn`, `specialist_consult`, `tool_invocation`, `scheduled_action`, `alert`, `integration_inbound` (Slack / Discord / email / Telegram / Google Chat), plus `knowledge_retrieval`, `fact_confirmation` (an emailed standing-fact change held, applied, cancelled or refused; private), `fact_reviewed` (the principal approved or declined a teammate's proposed standing fact; private), `fact_approval_changed` (the principal turned \"needs my approval\" on or off for a teammate; private), `fact_retired` (the principal, or the teammate who recorded it, retired a standing fact from the Pulse page; private, since it names the fact a teammate can no longer read; the fact tools' own rows and their `skill:` dispatch rows are private `tool_invocation` rows, since they quote the principal and name their session and turn), `memory_extraction` (one row per episodic-extraction pass, written in a `finally` so a crashed pass is recorded too: how many items the model proposed, how many were stored, a `{kind, reason}` record per drop, a `malformed` count for payload shapes that were never usable items, and a `failure` naming the exception type when the pass raised — including `CancelledError`, which is a `BaseException` and so would otherwise be logged as a clean empty pass — a `proposed>0, stored=0` streak means the quote validator is rejecting everything, which is otherwise invisible), `cache_event` (one row per model call from every call site — Executive turns, specialist consults (chat turns and workflow steps), research specialists, the research routing and watchlist passes, triage, memory extraction, alert review, the brief's narrative and its grounding repair, the digest, reflection and weekly-review workflows, committee reviewers, and the small utility calls (session titles, the chat-channel response gate, reply parsing, PDF transcription); a unit test fails any model call in the `openexecutive` package that records no row — carrying tokens, cache hits, cost, server-side search count and, inside a research run, its `run_id`), `memory_snapshot`, `committee_review`, `peer_memory`, `attunement` (one row per open-loop extraction pass — opened / closed / dropped with a reason per drop and a `failure` type — plus a row per loop opened or closed and per 👍/👎 on a reply; see Attunement), `executive_paused` / `executive_resumed` (the operator pause switch; actor is the caller's email, details carry the reason, or on resume the held-action count), `scheduled_action_cancelled` (a pending scheduled action cancelled from the Pulse page or a script; actor is the signed-in email, `admin-token` or `local`), `workspace_settings_changed` (solo/team mode, the user's time zone or the principal's role changed through `PUT /workspace`; actor is the caller's email or `api`, details carry the mode's and zone's from/to, plus `role_fields_changed` — the names of the role fields that changed, never their text — when the role changed, and the new `company_domains` list when the company's email domains changed), `decision_class_mode_changed` (a decision class switched between `propose` and `auto_execute` through `PUT /decisions/classes/meeting_scheduling`; same actor rule, details carry the class and the mode's from/to), the Act as me events — `delegation_settings_changed` (a person turned it on or off for themselves), `delegation_reminder_set` and `delegation_reminder_sent` (a person set a reminder for themselves with `remind_me`, and it went to them; ids and times only, never the text; private to that person), `delegation_actions_proposed`, `delegation_actions_approved` and `delegation_actions_dismissed` (the Executive left a person an approval card, which they approved or dismissed; kinds and outcomes only, never the text; private to that person), `delegation_gmail_verified` (their own Gmail checked before turning it on), `delegation_voice_changed` (\"How I write\" learned, edited, locked or reset, or its signature read from Gmail again), `delegation_voice_described` (a style written from the person's own description, counts only; nothing saved until they save it as an edit), `delegation_drafted` (one per draft saved in their Gmail from chat), `delegation_inbox_changed` (Draft replies to my inbox turned on or off), `delegation_inbox_scanned` (one per inbox scan, with its counts and status), `delegation_reply_drafted` (the inbox watcher saved a reply and made its card), `delegation_reply_dismissed` (a card dismissed, and whether its draft was deleted or kept), `delegation_reply_closed` (a card Gmail already settled: sent or deleted there, answered, gone or expired) and `delegation_reply_sent` (the person tapped Send and the draft went, or the reconciler confirmed an unconfirmed send did) — always private and metadata only: ids, counts and flags, never the draft or the thread; the roster-request events (see Org) — `roster_request_created`, `roster_ack_sent`, `roster_ack_withheld` (an acknowledgement the adapter declined — an email Gmail did not authenticate), `roster_request_resolved` (actor is where it was answered: `web`, a chat channel, `email` or `people_page`), `roster_message_replayed` (one per held message: replayed, dropped, failed or unavailable) and `roster_email_answer_refused` (an emailed answer with no live token, or failing DMARC) — always private to the principal and never carrying what the new sender wrote, like the `integration_outbound` row (`kind: roster_ack`) for the one gateway send allowed to an address off the roster, the fixed acknowledgement; `grounding` (unattended prose that named people or figures its inputs don't hold — a brief or digest that held back lines or used its repair call, a reflection tool call refused or flag held back, a triage or review rewrite replaced; details carry the surface, the mode, the ungrounded names and figures and up to five held-back lines cut to 200 chars, private when the brief read the principal's private rows or the alert is private; in `GROUNDING_CHECKS=report` mode these rows are written and nothing is changed); and the external-signal/monitor events. An unknown chat sender's `integration_inbound` row still reads `rejected_unknown_sender` — it means the message did not reach the Executive, not that it was discarded: a DM or mention is held as a roster request.\n\n### Session & turn context\nEach row carries `session_id`, `turn_id`, `actor`, `summary`, `details_json`, `full_json`, `department`, `private_to_principal` and `private_to_person`. A `turn_id` (`f\"t-{uuid12}\"`) is set per Executive turn and propagated automatically via ContextVars to retriever / specialist / tool / cache emits, grouping one user-assistant exchange. `summary` is the searchable column (LIKE, ~300 char cap); `details_json` is structured and not searched. `department` is filterable in SQL (`AuditLogger.query(department=…)`) — the department check-in's skip rule uses it to ask whether anything was logged for a department since its Goals were last reviewed.\n\n### Rows private to the principal\nThe principal reads `/audit`, and others their own rows (see below), so a row about something private to the principal is marked `private_to_principal` as it is written, and only the principal reads it. `AuditLogger.log` marks a row private when its caller passes `private=True` — the email poller's two inbound rows for mail from one of the principal's contacts or mail they forwarded, and `create_alert`'s row for a private alert — or when `people_tools.audit_row_private_to_principal` says so at write time:\n- every row a turn private to the principal writes (`Session.private_to_principal`): its reply, each tool call, the model-call rows, and the background passes that inherit the turn;\n- on the principal's own verified turn (or inside `grant_contact_egress`), any row whose summary, details or full payload names one of their contacts — their email address or full name anywhere in the text, or their person id or chat id in a person or recipient field (`person_id`, `attendee_person_ids`, `user_id`, `chat_id`, `channel_ref`), JSON tool results included.\n- every row an Act as me turn writes once `ghostwrite_email` or a mailbox read (`search_my_email`, `read_my_email`, `read_my_email_attachment`, `my_email_awaiting_reply`) has touched the speaker's mailbox (`TurnDelegation.touched_mail`, set before the tool's first read): the tool row, the reply's `chat_turn` row carrying the draft's preview, the model-call rows, and anything it starts in the background, under `private_rows` (it runs no learning passes). Only the principal can have Act as me for now, so private to the principal is the right visibility; the tool's input and result are also redacted (`ghostwrite` and `my_email` are sensitive tool names).\n- every row the Act as me inbox watcher, a reply card's Send or Dismiss, an action card's Approve (or the card carried out on its own, in training), or a person's Act as me settings write: they run under `rows_for_person(person_id)` (or pass `private_to_person`), so their model-call rows and events are private, and the reply cards are their own person's alone. That includes a person changing the AI line on replies sent as them (`delegation_disclosure_changed`) and what training records: `delegation_training_changed`, `delegation_reply_allowed`, `delegation_example_kept`, `delegation_actions_allowed`, `delegation_actions_handled`, `delegation_actions_held` and `delegation_learned_removed`.\n\n### Rows that belong to one team member\nAct as me can be on for team members as well as the principal (see Act as me). What a team member's own mailbox touches must not reach the principal either, so such a row also carries `private_to_person` — the team member's id — as it is written. A teammate's own conversation with the Executive is theirs the same way, as their mail is: every row of a turn a team member other than the principal speaks on a verified surface (the web app signed in, their own Slack or Discord, a secret-verified private Telegram chat) or by email Gmail authenticated as theirs (`people_tools.teammate_turn_owner`), its reply, tool calls, model calls and memory snapshot included. The rows written before the turn binds its session follow the same rule: the chat adapters' `audit_rows_on_senders_turn` holds `rows_for_person` for a verified teammate (`sent_by_teammate`), the web chat route for a signed-in teammate, and the email poller for authenticated teammate mail, which it checks before writing its inbound row. An unproven sender (a forged From, an unverified Telegram chat) keeps the ordinary rule, so nobody can hide a turn from the principal by claiming to be a teammate. `people_tools.audit_row_owner` decides it: the id a caller passes, else the `rows_for_person` scope, else the speaker of a turn whose `touched_mail` is set, else the teammate whose own turn it is; the principal's own id is never stored, so their rows stay private to the principal exactly as above, and a principal lookup that fails keeps the owner. A row with an owner is always private, and `AuditLogger.query` and `count` leave it out in SQL unless the caller asks for that person's rows (`owned_by`; `only_owned` keeps those alone, as the read routes do for a teammate) — so the brief's handled-overnight block and every other internal reader skip it. The read routes show it to the principal as an action only (`api.routes.audit._redact`): the summary is rebuilt from the event type and a plain tool name with whose conversation it was (\"mcp:send_gmail_message (Ben's conversation, details private)\"), the details keep only what was done and what it cost (`tool`, `kind`, `ok`, `refused`, token counts, …), and `full` is dropped; a text search (`q`) leaves such rows out, so their words never decide a match. The person it belongs to reads it in full. Only `usage_summary` counts it, so spend stays whole. `usage_summary` also adds up the sandboxed scripts (`run_script`) from their own rows — chat's `tool_invocation` with `kind: script` and a workflow step's `workflow_tool_call` with `tool: run_script`, each carrying `calls` and `duration_ms` — into `scripts` (scripts run and worked, tool calls, time, and `turns_avoided`, an upper bound on the model turns those calls would otherwise have needed); a script's individual calls are not counted twice. It also sums the `python_job` rows (one per `run_python_job`: `ok`, `duration_ms`, `cpu_ms`, `peak_mb`, files and bytes in and out, `attachments`, `saved_tool`) into `python_jobs`, with the `cache_event` usage of the turns that ran a job, each turn counted once. Rows that quote someone's own documents follow the same `rows_for_person` scope: a `draft_artifact` / `get_artifact` / `list_artifacts` dispatch row, the document tools' own rows, and a turn's `memory_snapshot` when its retrieval recalled a published document are private to the speaker (to the principal when they are the speaker).\n\nSome rows are written before their turn binds its session, so two scopes in `audit.context` count too. The email poller holds `private_rows` for the whole handling of a mail from a contact or a principal forward: every row is private, the knowledge retrieval (which quotes the mail) included. The chat adapters hold `principal_turn_rows` once they know the principal sent the message on a verified surface (`people_tools.audit_rows_on_senders_turn` on Slack, Discord and Telegram; the web chat route checks its caller), so the inbound row, the knowledge retrieval, alert triage and the start of the web stream follow the principal-turn rule above. Tasks started inside inherit the scope, except unattended work: a workflow run a chat reply resumes and each scheduled action start with neither (`unscoped_audit_rows`), so their rows follow the ordinary rule. A sender lookup that fails counts as not the principal, as the turn's own check does. Neither scope lets the turn reach anyone. A caller can add privacy but never remove it, and a check that cannot run keeps the row private. On anyone else's turn a contact is a stranger, so a row that names one keeps its visibility: hiding it would tell whoever wrote it that the address belongs to a contact. The poller handles each message with no session bound, so one message's private turn never marks the next message's rows.\n\nEvery read route resolves the caller as `/alerts` and `/decisions` do (`people.caller_is_principal`: the principal's signed-in email, or no `x-caller-email` — the CLI, direct curl, local login — once a principal is on the roster; before anyone is principal, nobody), in `api.routes.audit._view`. The principal reads every row, a teammate's own as an action only (above). Once a principal is on the roster, anyone else reads their own rows (`owned_by` = them, `only_owned`) and nothing more, since the log holds what everyone said to the Executive; a caller the roster can't resolve reads none. Before anyone is principal, everyone reads the rows not private to one. All of it is filtered in SQL, so `total` and pagination stay honest: `GET /audit/logs` leaves the rest out, `GET /audit/logs/{id}` answers 404 exactly as for a missing id, and `GET /audit/sessions/{id}` shows only the rows the caller reads (404 when none are left). `GET /audit/usage` is spend, not content: for anyone but the principal it leaves out the rows private to the principal, as before. Internal readers see every row unless they ask otherwise; the department check-in, whose report goes to the department, reads its recent activity without them, and its \"nothing new since the review\" skip rule ignores them too. Ids are one autoincrement sequence, so anyone else can still see a gap where a private row was written; that is accepted.\n\nAlways in the loop writes `history_notes_added` (count and source, never the note) and `history_settings_changed`, `history_note_changed`, `history_note_forgotten` and `history_conversation_forgotten` from its routes. Every one is private and belongs to its person alone (`private_to_person`). `recall_history` is a sensitive tool name, so its input and result are redacted.\n\n### Views\n- `GET /audit` — filterable list with a \"Group by session\" toggle and session pills.\n- `GET /audit/session/{id}` — full ordered timeline plus a server-derived node/edge graph rendered as a three-pane React Flow chart.\n\n### Invariant\nAudit writes never break a chat turn: all writes swallow exceptions and log a warning, and instrumentation observes only — it never touches request building or the cached prompt blocks. Audit is append-only: no UPDATE or DELETE paths exist in product code.",
"mermaid": "flowchart TD\n Chat[\"Chat turn / consult / tool call\"]\n Sched[\"Scheduled action\"]\n Inbound[\"Inbound integration\"]\n Log[\"audit/logger.log_event\"]\n Priv[\"private_to_principal? private turn, or names a contact on the principal's turn\"]\n DB[\"audit_log table (append-only)\"]\n Gate[\"read routes: private rows for the principal only\"]\n ListView[\"/audit list view\"]\n SessView[\"/audit/session/{id} flow chart\"]\n Chat -->|event + turn_id| Log\n Sched --> Log\n Inbound --> Log\n Log --> Priv\n Priv -->|swallow errors, never block| DB\n DB --> Gate\n Gate --> ListView\n Gate --> SessView\n class Chat entry\n class Sched entry\n class Inbound entry\n class Log compute\n class Priv compute\n class DB storage\n class Gate compute\n class ListView hot\n class SessView hot",
- "generated_at": "2026-10-12T13:58:17+00:00"
+ "generated_at": "2026-10-13T15:30:00Z"
}
diff --git a/packages/core/openexecutive/architecture/prebuilt/delegation.json b/packages/core/openexecutive/architecture/prebuilt/delegation.json
index 3a0afb77..5d41118d 100644
--- a/packages/core/openexecutive/architecture/prebuilt/delegation.json
+++ b/packages/core/openexecutive/architecture/prebuilt/delegation.json
@@ -1,7 +1,7 @@
{
"section_id": "delegation",
"title": "Act as Me (Delegation)",
- "markdown": "The **delegation** module (`openexecutive.delegation`) is **Act as me**: the one place the Executive writes *as* a person instead of as itself. Everywhere else it has its own Google account and bots and signs as itself. Here — when the person asks, or, with a second switch, for mail in their inbox that needs them — it writes an email in their voice and saves it as a **draft in their own mailbox** — Gmail or Outlook — for them to review and send. The one thing it ever sends is one of those inbox drafts, exactly as it is in their mailbox: when the person taps **Send** on its card, or, with a third switch, **Handle it for me**, on its own when plain-code rules allow it (below); chat never sends. The principal can turn it on for themselves, and so can each team member once the owner lets them. It is off by default, and off changes nothing: the prompts, tool lists and prompt caches are byte-identical to an install without it.\n\n## Who can have it, and when it is offered\n\n`delegation.settings.can_delegate(person)` is the single rule for who may turn it on: the principal (a non-archived team member flagged `is_principal`), and any other non-archived team member — never a contact — once team members may have it (`team_members_enabled`). That takes two things: the install allows it (`DELEGATION_TEAM_MEMBERS`, default off, which offers the owner the switch) and the owner turned on **Let team members use it** (`PUT /delegation/team`, principal only, one row in `delegation_team`). Turning the owner's switch off takes it away from every team member from their next message, and their inbox watchers stop at their next check; their own switches, drafts and cards are kept. Everything is keyed by `person_id`. A person turns it on for themselves (`PUT /delegation`); nobody turns it on for anyone else. The setting is one row per person in `delegation_settings`, and a missing, unreadable or zero row all mean off.\n\nAt the start of every chat turn (`Executive.stream_chat` and the committee path) `pin_turn_delegation(session, speaker_text)` runs next to the workspace-mode pin and records a `TurnDelegation` on the session. `ghostwrite_email` is **offered** only when all of these hold:\n\n- the speaker may have it and turned it on — read fresh every turn, so turning it off applies from the next message;\n- the speaker is on a surface that verified it is them — the principal through `people_tools.is_principal_on_verified_surface`, a team member through `teammate_on_verified_surface` (the web app signed in, their own Slack or Discord, never an email turn) — and a web turn also carried a signed-in caller (`Session.web_caller_signed_in`, set from `x-caller-email`) or runs under local login — a request with no caller header is not a sign-in and never gets it (whoever holds the shared secret is trusted as the UI proxy that stamps that header, as on every principal-only route);\n- the conversation is private to the speaker: the web chat, a Slack or Discord DM, or a verified private Telegram chat — never a shared channel or thread, where the draft's preview or the matching threads would be posted for everyone and other people's messages sit in the model's context;\n- the turn is neither unattended (the scheduler's proactive run) nor private to the principal (the email poller's turns), so inbound text can never reach it. The MCP server, workflows, reflection and research never carry it at all.\n\nThe mailbox connection is deliberately **not** part of the offer: the handler checks it on every call and says how to fix it, so the cached tool prefix never flips when a token lapses. The pin lives in a context variable for the turn's task as well as on the session, so two turns running at once on one session (a second browser tab) each read their own. It also carries the speaker's own words for the turn (a new email may go to an address they just typed — the turn's message joins the history only once the turn ends), a per-turn draft count, and `touched_mail` (see Privacy). `own_words` reads those words back for every own-words check (addresses here, facts and task assignments elsewhere): it first removes the hydrated `<outbound_reply_context>` backstory by its exact, anchored shape, so a look-alike closing tag quoted inside it cannot pass the rest off as the speaker's, then drops other adapter tag blocks, and returns nothing for a message carrying an attachment.\n\n## The tool: `ghostwrite_email`\n\nIt lives in its own registry (`orchestrator.delegation_tools.DELEGATION_TOOLS` / `DELEGATION_TOOL_HANDLERS`), **never** `_ALL_SKILL_TOOLS`, so no other toolkit can carry it. On a turn it is offered to, the agent loop adds it to the pool *before* the sort and builds that turn's handler map from both registries; on any other turn a call to it is an unknown tool. The handler re-checks the pin and the surface anyway, then:\n\n1. **Caps** — at most 5 drafts per turn, and `DELEGATION_MAX_DRAFTS_PER_DAY` (default 50, at most 1000) per person per UTC day, shared with the inbox watcher (`delegation.caps`). It is counted from `delegation_drafts`, one row of ids per draft saved from either place, with a count kept in the process as a floor in case a row was lost. Each slot is taken before the first await — a round's tool calls run concurrently, so otherwise every call in it would pass the caps before any counted — and handed back if nothing was drafted. If the rows can't be read it refuses rather than guess.\n2. **Marks the turn** — sets `TurnDelegation.touched_mail` *before* its first read of the mailbox.\n3. **Checks the mailbox** — `gmail_status` asks Google or Microsoft whose mailbox the credential opens and compares it with the person's People email, on every call: a roster email can change, and a client slot swaps the roster.\n4. **Finds the thread** — by `thread_id` (checked against the mailbox's own id shape), or a Gmail-style search (`find`; on Outlook, words, `from:`, `to:`, `cc:`, `subject:` and `newer_than:` are kept and other Gmail operators dropped). Several matches return `{status: \"choose\", candidates}` (subject, sender and date — no message text) and draft nothing; none returns `not_found`.\n5. **Sets the recipients in code**, never from the model. A reply goes to the last inbound message's `From` — never its `Reply-To` (a different one is flagged `reply_to_ignored`) — with the thread's other recipients only on `reply_all`. A new email or a forward may go to any well-formed address: it is only a draft, waiting in their own mailbox, where they see who it is to before they send it. An address that is neither on the roster (contacts only when `contacts_reachable_now()`) nor typed by the speaker this turn comes back in `not_in_people`, and the Executive names it for them to check (`delegation_tools.not_in_people`). Mailing-list mail, a thread they replied to last and a reply-all trimmed to 10 recipients are flagged. An address inside a block an adapter added to their message (the backstory hydrated into a DM) is not one they typed, and when the message carries an attachment — a document's inlined text, or a note naming the file — none of its addresses count as typed: neither has an end mark, and a filename is chosen by whoever sent the file, so such an address is named back to check.\n6. **Writes it** (the ghostwriter, below) and saves it with the mailbox client's `create_draft` (Gmail: threaded with `In-Reply-To`, `References` (whole message ids: when the chain is too long for one header the oldest after the thread's first go, never the parent) and `threadId`, and stamped `X-OE-Ghostwritten`; Outlook: Graph's `createReply` on the message it answers, then the text and recipients, and a named property marking it).\n\nThe result is `{status: \"drafted\", draft_id, gmail_link, to, cc, subject, preview, flags, open_questions, note}`, plus `not_in_people` and `check_address` when a recipient needs checking. The link (`gmail.mailbox_link`) is built server-side from a fixed prefix — the thread in Gmail, the draft itself in Outlook on the web — and the note tells the model the preview is the person's draft, to be treated as data. The chat shows an action chip — \"Drafted an email as you to … — in your Drafts\" — that opens it, and only a link under `MAILBOX_LINK_PREFIXES` is ever linked. A sincere \"are you an AI?\" in the thread comes back as an open question for the person to answer themselves.\n\n## Reading their mailbox\n\nFive read-only tools ride with `ghostwrite_email` in the same registry (`orchestrator.mail_read_tools`), so they are offered, re-checked and fenced exactly as it is: `search_my_email` (a Gmail-style search through the client's `search_threads`, or with no query the threads that reached their inbox in the last few days through `inbox_message_ids` + `get_message`), `read_my_email` (one thread's last 10 messages, each sender's new text only, drafts left out, with the files attached to them listed by message id and index), `read_my_email_attachment` (one of those files as text: PDF, Word, Excel, CSV or text, up to 20 MB, a scanned PDF converted) and `my_email_awaiting_reply` (their own sent mail from the last 14 days, to at most 10 people besides the Executive, that is still the last message in its thread). They change, label, draft and send nothing.\n\nThe fifth, `my_email_read_before`, lists the threads `read_my_email` opened for the same person before, in this conversation or an earlier one, so a later conversation can open the same email again instead of asking them to search for it. It keeps where, never what (`delegation.mail_reads`, table `delegation_mail_read`): the thread id, its subject and first sender as one line each, the mailbox address it was read from and when, never a message's text, an attachment or a recipient. Rows are the person's own and are listed only from the mailbox open now, newest first, optionally narrowed by words in the subject or sender; anything older than 30 days is never listed and is deleted on the next write, at most 200 are kept per person, and turning Act as me off deletes them all. It opens the mailbox like the other reads, so it sets `touched_mail`, counts as a search, and returns its rows with the same data-not-instructions note. The Act as me addendum tells the model that these tools reach the person's mailbox while any Gmail or Outlook tools of its own account reach only the Executive's, to search theirs (never falling back to its own) for mail sent to them, and to reopen an email from before with `my_email_read_before` and `read_my_email`.\n\nEach sets `touched_mail` and marks the conversation theirs before its first read (`_open_mailbox`, shared with `ghostwrite_email`), so the lockdown below applies from its round on. Each is capped per turn on the pin: 10 searches (`my_email_awaiting_reply` counts as one), 20 thread reads and 5 attachment reads, the slot taken before the first await. In `read_my_email` every message someone else wrote comes back inside its own `<untrusted_content source=\"email\">` block (`content_trust.wrap_untrusted`, which defuses a closing tag in the text); the speaker's own messages — SENT and from their address (`threads.mine`), never a From header alone — are marked \"Yours\" outside any block. An attachment's text comes back in its own `<untrusted_content source=\"attachment\">` block (`attachments.format_attached_text`), parsed in an isolated worker (`knowledge.loader`); unlike a file sent to the Executive it is never added to the company's knowledge. The clients find an attachment by message and position and look its provider id up on each read (Gmail's changes per fetch): Gmail lists a message's files with the message (`MailMessage.attachments`, leaving out images the HTML shows in place), Outlook says only `hasAttachments` until `list_attachments` asks Graph (leaving out inline ones), and `attachment_bytes` reads one; an attached Outlook item or cloud link has no file to read. Results carry a note that subjects, senders and text are other people's words, data rather than instructions. The lockdown refuses tools, not the reply's own text, so what would carry that text out without a click is closed where it is shown: the web chat loads no image from another site, and bot replies go out without link previews (see Integrations).\n\n## Forwarding and reminders\n\n`ghostwrite_email` also forwards: with `forward` (a thread id) and `to`, it drafts a forward of the thread's latest message, `intent` becoming the short note above it (the ghostwriter writes only the note; the forwarded email's subject, someone else's words, is never put in the intent). Recipients follow the new-email rule (`_new_recipients`: the roster, or addresses the speaker typed this turn). Gmail's draft quotes a forward header and the original's own words under the note and attaches the original's files, up to 10 and 20 MB (`FORWARD_MAX_FILES`, `FORWARD_MAX_BYTES`; any left off are flagged `attachments_left_off`); Outlook uses Graph's `createForward`, which keeps the original and its files, with the note as its comment and the body never patched. It is still only a draft.\n\n`remind_me(text, when)` (`orchestrator.reminder_tools`) rides with the same tools and checks, but opens no mailbox, so it never marks or locks a turn (`MAILBOX_TOOL_NAMES` are the ones that do). It stores the text once cleaned — one line, anything a chat app could turn into a link (a URL of any scheme, an address, a domain-like word, a host and port, an IP address) replaced with \"[link]\", at most 300 characters — with its time (stored in UTC) in `delegation_reminders` (5 a turn, 25 waiting, a year ahead, refused when the person has no channel). When it is due the scheduler sends that text to the person alone, as \"Reminder you set on <date>: …\", running nothing (see Scheduler); the Slack and Discord DMs it may go by turn link previews off. The chat chip shows the text as it will be sent. A reminder that comes due in the person's quiet hours or leave (when `OUTBOUND_RESPECT_QUIET_HOURS` is on) waits unclaimed and goes once they end, rather than being suppressed and lost. It is allowed after mail is read and in later turns of that conversation: it reaches only the speaker, as plain text. `schedule_followup`, which runs a turn with tools and is listed to everyone, stays refused there.\n\n## Approval cards\n\nA turn that read someone's mail may not message anyone, invite anyone or change the roster (see the lockdown above), yet what the mail asks for is often right. `propose_actions` (`orchestrator.action_card_tools`) leaves the speaker a card instead: up to five actions, each spelled out exactly as it will happen, done only when they tap Approve. It rides with the other Act as me tools and checks (`_writer`), opens no mailbox, and is allowed after mail is read because it does nothing on the turn.\n\n- **Kinds.** `invite` (a meeting with people on the roster), `message` (a direct message to someone on the roster, as written) and `add_contact` (a new contact; the principal's cards only, since contacts are theirs). A team member's card can't name a contact. Each action is checked when the card is made (`delegation.action_cards.check`) and again when it is carried out, so an archived person or a time that has passed fails that action alone.\n- **Whose.** A card is a `delegation_actions` decision for the speaker, `approver_only` like a reply card: only they see it (`GET /delegation/actions`) or resolve it, a team member's answers 404 to the principal too, and it is never an alert. The same actions still waiting are that card, not a new one. 3 cards a turn, 25 waiting, dropped after 7 days.\n- **Approve.** In the chat, under the reply that left it (the `propose_actions` chip carries the card's `decision_id`, and `ChatActionCards` shows the card while it waits), and on the Briefing's Needs you list, with a box per action. Approve (`POST /decisions/{id}/approve`, `edits.only`) needs the card's person signed in (`verified.caller_refusal`) with Act as me on; the card is claimed first so it runs once, then each ticked action runs as stored with no model call, in an unattended session (the principal's with contact egress) under `rows_for_person`, so every audit row it writes is theirs and nothing reaches the team's activity rail: `message_person` (a direct message or nothing; never its Briefing-alert fallback, which more people read), `create_calendar_event`, or a contact row. Approving also needs the person still allowed Act as me (`can_delegate`) and switched on. An invite is booked straight away when the person approving is the one the meeting gate would ask (`calendar_tools.approved_by`), and otherwise goes to that approver as usual, so a card never widens anyone's authority. Per-action results (`done`, `waiting`, `failed`, `skipped`) are kept on the decision and shown on the card. Dismiss rejects it and does nothing else. With Suggested actions in training the card adds Approve + allow, and a card of actions all allowed is carried out on its own (see In training above).\n- **Channels.** A card can come from any surface Act as me runs on; it is approved in the web app, since the chat apps have no buttons here.\n\n## The ghostwriter\n\n`delegation.ghostwriter.compose` writes the text. The Executive decides *what* to say (`intent`); this call decides *how* the person would say it, and it is deliberately small:\n\n- a constant `GHOSTWRITER_PROMPT` plus the person's voice block as the system prompt;\n- the thread (each message cut to its sender's own words, every line scrubbed, all labelled alike), then the writer's own earlier words in it (`<writer_said>`, from their sent mail only), the intent and the recipients' names in the user turn — no company profile, memory or roster;\n- with Always in the loop on, `<writer_noted>`: what the writer's own notes say they told these same people in earlier Act as me replies (`memory.history_drafts.notes_for_draft`: only notes from mail **every** recipient of the new draft was on, so nothing told to one person reaches another; chat notes name no recipient and never qualify; at most 8, newest first, the person's own words or their correction, with dates). The prompt's rule 6 makes it a consistency check, never content: a draft that contradicts it is still written as the intent says, and the conflict becomes an open question on the card, which only the writer sees. Both the chat tool and the inbox watcher pass it;\n- **no tools** but the forced `compose_reply` return, so text in the thread can ask for anything and there is nothing to do it with.\n\nIt is not a Council agent — the Council's edit routes have no principal check — and runs on `DELEGATION_COMPOSER_MODEL` (default: the default model), logged as `ghostwriter` usage. `lint` then runs in code on what comes back: a link or an email address survives only if it appears in the intent, the signature or the recipients, so a link planted in the thread is removed and flagged; the Executive's own name is flagged; and the body is capped, cut at a paragraph, line or word break in its last fifth. A draft with nothing left after that is refused. The signature from the person's Gmail settings (or the one already stored, for Outlook) is appended.\n\n## How I write\n\n`delegation.voice` learns how a person writes *their own* email (`attunement.style` learns how the Executive should write *to* someone). `POST /delegation/voice/learn` reads up to 40 recent sent messages, keeps only what they wrote (quoted replies stripped with the email poller's `sender_new_text`), skips automatic mail, every thread chat drafted into, and every sent message that began as a draft written for them (`delegation_drafts.sent_message_id`), and makes one forced-tool call (`record_voice_profile`). Code validates every field before it is stored:\n\n- greetings per audience (their team, their contacts, anyone else), a sign-off of at most two lines, and a usual length and tone (a backslash-n the model wrote instead of a line break is read as one, keeping the first lines that fit, which also repairs a profile stored before that rule);\n- up to 8 habits and 6 things to avoid, through the same deny-list as working-style rules — no actions, links, handles, amounts, tool names or names of people;\n- up to 3 short exemplars, each checked to be verbatim from a sample, with addresses and phone numbers masked — and none while a client slot is active, so one client's mail never lands in another's state.\n\nThe signature comes from Gmail's `sendAs` settings, not the model (Outlook gives apps no way to read one, so a learn keeps the stored signature and `POST /delegation/voice/signature` answers 409 `signature_unavailable` rather than clearing it), as plain text with its lines kept (`gmail.html_to_text`: a `<br>` always ends a line, and a block tag, opening or closing, ends one when it has text, so Gmail's layout of a first line followed by one `<div>` per line comes out line by line). `POST /delegation/voice/signature` reads it again without relearning: nothing else changes and a locked profile stays locked, so it picks up a changed Gmail signature and undoes **Don't add my signature**. A person can also describe their style in their own words: `POST /delegation/voice/describe` applies the description to the profile they have through the same forced tool (a separate system prompt, no exemplars) and the same validation, keeps their signature and writing examples, and returns the result with a short sample reply for them to review. It saves nothing and needs no mailbox; Save stores it as an ordinary edit. The person can edit the profile, lock it (a locked profile is never relearned) and reset it, and every change is kept in `delegation_voice_history`. One learn runs at a time per person, and a lock or an edit made while it runs wins: the learned profile is dropped (`changed`).\n\n## Draft replies to my inbox\n\nA second switch under Act as me, **Draft replies to my inbox** (`delegation.inbox`), has the Executive watch the person's own inbox and write a first reply to mail that needs them. The reply waits on a private card on Today; nothing is sent until they tap **Send** (below). Where the reply lives is the person's choice, **Also save replies in my Gmail Drafts** (`mailbox_drafts`, off by default): off, it lives on the card alone (`delegation.card_drafts`: the card keeps the draft's recipients, subject and thread headers as `draft_spec`, and Send makes the mailbox draft from exactly that just before the send-time checks, taking it back out if they refuse), so their Drafts folder doesn't fill with replies they may never send; on, it is also saved in their Drafts as before. The switch is one row per person in `delegation_inbox_watch`. Turning it on needs Act as me on and their mailbox connected, and sets `watch_since` to now every time, so it never drafts for mail that arrived before; turning Act as me off turns it off too. Off means no mailbox call and no row written, with one exception: a send the person started that the mailbox never confirmed is still followed through until it is settled.\n\n### When it looks\n\nA throttled hook in the scheduler tick (`_maybe_scan_inbox`, after the pause, company-profile and rotation gates) checks each person whose switch is on every `DELEGATION_INBOX_POLL_MINUTES` (default 5). It is not a scheduled action, because those are listed on `/scheduled` for everyone and anyone may cancel them. One scan task runs at a time, one scan per person, and none while a client slot is active. **Check now** on the Settings card runs the same scan. Every row a scan writes is private and the person's own (`audit.rows_for_person`: a team member's rows are hidden from the principal too, see Privacy), and the process log carries counts and codes, never an address or a line of mail. The scan never marks mail read or changes a label.\n\n### What it drafts for\n\nA scan first settles the open cards (below), and stops with `backlog_full` once 25 are waiting, counting the ones it makes. It then lists the inbox since `watch_since` (`inbox_message_ids`) — not chats, not mail from the person, and not the promotions, social, updates or forums tabs (on Outlook: the Focused inbox only) — up to 100 messages, and takes the newest message of each thread that still needs a look, at most 10: settled threads and threads with an open card are passed over first, so a burst of mail, or a pause for a limit or a back-off, never hides older threads for good. `delegation_inbox_messages` is its durable cursor: one row per message (`UNIQUE(person_id, message_id)`) with what happened to it, the reason, how many tries it took, and the draft and card ids. It stores a hash of the sender's address, never the address. A message is skipped, with the reason recorded, when:\n\n- it came before `watch_since`, or more than three days ago;\n- it is from the person or one of their send-as addresses, from the Executive, or one of this module's own drafts;\n- it is automatic — `Auto-Submitted`, a bulk or auto-reply `Precedence`, a bounce, a calendar invite or a no-reply sender — or from a mailing list, or in a Gmail tab, spam or trash;\n- the person is not in To or Cc, or it has more than 10 recipients;\n- the person replied last, or a draft already sits in the thread.\n\nThree kinds are left unrecorded, so a later scan looks again: mail under 10 minutes old (mail the person answers straight away never gets a card), mail in a thread whose card is still open, and mail a daily limit holds back.\n\n### Deciding, then writing\n\nTwo model calls, and code decides between them. The classifier (`delegation.inbox_classifier`) is one forced-tool call, `classify_email`, returning whether the email needs a personal reply, what kind it is and how sure it is. It sees a few one-lined header lines and the sender's own new words (at most 3000 characters, every line scrubbed) inside one `<email>` block marked as data, and it has no other tools, context or memory, so an email can ask for anything and there is nothing to do it with. A reply is drafted only for a question, a request, scheduling, an introduction or a follow-up, and only when the confidence clears a bar that rises the less the sender is known: 0.6 for their team or a contact (the principal's contacts count only for the principal: a team member's mail is weighed by the team and their own sent mail), 0.7 for someone they have written to before (`has_written_to`, a search of their sent mail), 0.85 for a stranger. Pitches, notes of thanks, FYIs, notifications, newsletters and phishing get nothing, and so does any failure. An email that also went to other people is drafted for only when the person is in To and it asks them themselves: the classifier is told their name, whether they are in To or Cc and how many others it went to, and lists the names the email greets or asks by name (`asked_names`). Code matches those against the person's name, whole name parts only (\"Alexis\" never counts for \"Alex\", and a nickname misses, which means no draft), so mail that greets someone else (\"Good morning, Brennan, could you sign ...\") is never theirs, whatever the model makes of it as a whole. When it names nobody, the model's `asked_of_them` decides, false for a question to the group at large. One that only copies them is never drafted for. Either is recorded `not_needed` with the reason `asks_someone_else`. It runs on `DELEGATION_CLASSIFIER_MODEL`, else the routing model. A sender whose address Gmail couldn't authenticate is handled as a stranger whatever the roster says — anyone can put a colleague's or a client's address in From — so they get the stranger's bar, a holding reply and the stranger limits; the card still says who they claim to be, and that the address isn't verified.\n\nThe reply is written by the same ghostwriter, in the person's voice, but from a fixed intent the module builds rather than words the person typed. Because the intent lets it restate the person's own earlier words, which words are theirs is never written inside the thread, where a sender could imitate any marker with a look-alike: every message there is labelled alike, and the person's own words come in a separate `<writer_said>` block after the thread, taken only from mail their own mailbox sent. Nothing inside the thread can close the thread block or open another. The intent: acknowledge the sender; say again only what the person's *own* earlier messages in the thread already said; commit to nothing new — no yes or no, date, time, price, fact or promise — and say they will get back to them; and put every question it could not answer from their words in **open questions**. A stranger always gets a short holding reply. The lint still strips any link or address that isn't the person's own. The reply goes to the sender (never `Reply-To`) and, on a group email, to everyone else it went to as Cc (reply to all, never the person's own addresses or the Executive); a stranger's holding reply goes to the sender alone. The card says when others were on the email, and its `draft_to` lists everyone, which Send checks.\n\nThe order: record the message, classify, reserve a slot in the shared daily limit, write, save the draft in Gmail (from the alias the mail went to, when it went to one) and store its id, then create the card (idempotent on the message) and mark it drafted. Nothing that fails along the way loses the message. A failure that may pass — the classifier or the writer failing, Gmail refusing the draft, a card that couldn't be made (its draft is deleted first, so no draft is left without a card), any other error with that one message — leaves it for a later scan (`retry`), and it is given up on only after three tries; one message's trouble never holds up the rest. A daily limit reached while it was being read leaves it for later without counting as a try.\n\n### The reply cards\n\nA card is a `delegation_reply` decision whose class is **its approver's alone** (`DecisionClassSpec.approver_only`): only the person it was drafted for sees or resolves it on any decisions route — a team member's card answers 404 to the principal too — and it is never an alert: alerts feed chat turns, which would put other people's mail into memory. It carries the sender with who they are to the person and whether Gmail could verify the address (SPF/DKIM/DMARC, through `fact_confirmation.headers_authenticated`), the subject, what they wrote, the draft, the open questions and its warnings. `GET /delegation/replies` lists the open ones from the database alone, so Today never waits on Google.\n\n- **Send** approves the card, which sends its draft exactly as it is in their mailbox (see *Send on your tap*). **Edit** changes the reply's words right on the card (`PUT /delegation/replies/{id}/text`, `card_drafts.edit`): only its own person, from a caller the API can tie to them (as for Send), never while it is being sent on its own; a reply in their Drafts is rewritten there too (`update_draft_text`, every header kept), and only while it is still the version the card last saw, so a change made in the mailbox is never overwritten. A card edit counts as edited (`approved_with_edit`, and **Do it like this next time** keeps it). **Open in Gmail** (or **Open in Outlook**), shown only when the reply is in their Drafts, opens it there.\n- **Dismiss** rejects the card; the class's after-reject step then deletes the draft if nobody edited it — its message id is still the one it was saved with, since Gmail gives an edited draft a new one, and the Outlook client makes a draft's message id `<id>.<changeKey>` so an edit there changes it too — and keeps it otherwise. What happened is recorded even when Gmail fails.\n- Each scan first **settles** what Gmail already settled: a draft sent from Gmail, a draft deleted there, a reply the person wrote themselves, a thread that is gone, or a card left for 7 days (its draft stays in Gmail). Each is a compare-and-set close with the reason. A newer message from someone else flags the card instead. A card being sent at that moment is left alone.\n\n### Send on your tap\n\n`delegation.reply_send` is the only thing in Act as me that sends, and all it can send is the draft on a card, by its id: `DelegateGmail.send_draft` posts `{\"id\": …}` to Gmail's `drafts.send`, and `DelegateOutlook.send_draft` posts an empty body to that draft's Graph `send`, and nothing else, so no text, recipient or header is ever passed. It has two doors: `send_approved_reply`, reached only from `POST /decisions/{id}/approve` — a class only the card's own person may resolve — and `send_on_its_own`, reached only from the inbox watcher under Handle it for me (below). AST tests pin both call sites, so no model, chat turn, workflow, scheduled job or MCP call can. A tap is checked like this: At send time it checks everything again:\n\n1. **Who is asking.** The caller is the card's person and the API knows it: signed callers are on and the request carries that person's signed assertion (or the operator's, under local login), or the API runs for local login and answers this computer only. Anything else is a 409 `caller_signing_required` — with only the shared secret, whoever holds it could send as the owner — so on a server Send needs signed sign-ins.\n2. **The switches.** Act as me and Draft replies to my inbox are on, and no client slot is active.\n3. **The mailbox.** Their mailbox is connected.\n4. **The draft.** It still exists, in the same thread, and they haven't replied there themselves since; otherwise the card closes.\n5. **Who it goes to.** To, Cc and Bcc hold between 1 and 10 addresses, never the Executive's own, and it is from one of their own addresses.\n6. **What changed.** Recipients that differ from the card's, or a newer message from someone else, need a second yes: the first tap gets a 409 `confirm` naming the recipients and the reason, and the person approves again with what they were shown.\n\nThen it claims the card (`claim_for_execution`, proposed → executing, a compare-and-set, so a double tap sends once) and reads the draft once more: `drafts.send` sends whatever the draft is when Gmail gets the request, so if it was edited since the checks read it — in another tab, by a delegate — nothing is sent and the card comes back (`draft_changed`). Then it sends, and records the result: `approved_with_edit` when the draft was edited in Gmail, the sent message's id, the draft row's `sent_message_id`, and a private `delegation_reply_sent` row. A refusal that proves nothing went — a 4xx, a lapsed sign-in, a rate limit, a draft already gone — hands the card back. A failure that leaves it unclear — a timeout, a 5xx — leaves it `executing` and says to check the Sent folder, and the next scans ask Gmail: their message in the thread means it was sent; anything else is judged on the second scan that sees it: the draft still there and nothing sent means it was not (Gmail deletes a draft it sends), and the card comes back, flagged; the draft gone and nothing sent means it was deleted. Nothing is ever retried on its own. Sending is the person's own act, so it works while the Executive is paused.\n\n### Limits and failures\n\nAt most 5 drafts a scan; `DELEGATION_INBOX_MAX_DRAFTS_PER_DAY` (default 20) a day, inside the shared `DELEGATION_MAX_DRAFTS_PER_DAY` it shares with chat; 200 classifications, 2 drafts per sender (1 for a stranger) and 10 for strangers in all per day; and never more than 25 open cards. A lapsed Gmail sign-in sets the status and is looked at again every 30 minutes; a rate limit, an error reaching Gmail or a crash backs off from 5 minutes, doubling up to 2 hours. Neither turns the switch off: the Settings card shows the status in plain words.\n\n## Handle it for me\n\nA third switch under Act as me, **Handle it for me** (`delegation.handle_it`), lets the inbox watcher send some of its replies itself instead of leaving them on a card. It is one row per person in `delegation_handle_it` (`enabled`, `mode`); no row means off. One setting says how much goes without them, the way quality does on the Agent Council (`handle_it.RULES`): **Easy ones** (`careful`) sends only replies to people they know (someone on their team, a contact or someone they have written to) of at most 400 characters when the classifier is at least 0.9 sure; **People I know** (`balanced`), where it starts, sends replies to people they know up to 1200 characters at 0.8; **Most mail** (`bold`) also answers strangers, up to 2000 characters, and lets a link or an amount through when the classifier is at least 0.9 sure. Everything else waits on a card as before, so turning it on never drafts less. A table from before the setting gets the `mode` column on its next open (`schema._add_handle_it_mode`): someone who had strangers set to send keeps that as Bold, everyone else gets Balanced. Only the person turns it on (`PUT /delegation/handle-it`), from a session the API knows is theirs: the same rule Send uses (`delegation.verified.caller_refusal`), so with only the shared secret it answers 409 `caller_signing_required`. Turning it on needs Act as me and Draft replies to my inbox on, and turning the inbox watcher off (or Act as me, which turns it off) turns it off too.\n\n**In training** (`delegation.training`). Like someone new, each Act as me job can be put in training on its own, from that job's card on the Act as me page, the way Take the lead has it: **Write drafts as me** and **Handle it for me** each pick Off / In training / On (Handle it's In training is Replies in training, with a *Train follow-ups too* box for Follow-ups), and **Suggested actions** picks Always ask / In training; one *What it's learned* list sits at the bottom, each item named by its job (`PUT /delegation/training`, one row per person and setting in `delegation_training`, absent means not in training). In training it asks first and learns who and how from what the person approves; what it learns is a `take_the_lead_allowed` row (feature `act_as_me`, the person's id), theirs alone.\n\n- **Replies.** Every reply waits on its card (`handle_it_reason` `training`) whatever the dial says, and a reply card adds **Send + allow**: it sends that reply on the person's own tap, exactly as Send does, then allows replies to that sender (key `act_as_me|person:<id>|reply|<address>`). While Handle it for me is on, a reply to an allowed sender goes on its own under the dial's limits (a stranger they allowed included) and every check above (the topics that always wait, links and amounts, new recipients, once a thread a day, the daily limit); it's checked again at send time, so a sender removed since waits. Take the lead as you lifts nothing in training: putting Replies or Follow-ups in training turns it off, and it can't be turned on while they are (409 `in_training`).\n- **Follow-ups.** Every follow-up waits (`follow_up_training`); Send + allow on a follow-up card allows follow-ups to exactly the people it went to (`…|follow_up|<sorted addresses>`), whoever they are, under every other follow-up check.\n- **Suggested actions.** An action card (below) adds **Approve + allow**: it does the actions on the tap, then allows each one that happened (a message to that person, `…|message|person:<id>`; a meeting with exactly those people, the person left out, `…|invite|people:<ids>`). A new contact is never allowed. A card proposed later whose every action is allowed is carried out at once, with no tap (`action_cards.run_on_its_own`, `gate_mode` `auto_execute`, status `executed`), only if plain code finds nothing against it (`on_its_own_refusal`): still in training, Act as me on, a server that ties the switch to the person, nothing sensitive, no link or amount, a message of at most 1200 characters, and at most 10 actions carried out this way a day (checked one card at a time); anything refused is handed back to wait on its card as before. An invite carried out this way is never booked as approved by the person: the meeting gate decides as for any meeting nobody tapped, so a time taken from someone's mail can't land on a calendar unasked. The tool result and chip say what was done. The message text in such an action is model-written from the mail and goes without anyone reading it; the allowance covers who and what kind, not the words, which is why sensitive topics, links, amounts and long messages are held back.\n- **Drafts.** Nothing is sent. When the person changes a reply's draft in their mailbox before tapping Send (plain or + allow) and ticks **Do it like this next time**, what they sent is kept (its new text, quoted replies dropped, at most 1000 characters, `…|style|<address>`), and the ghostwriter gets it in its user turn as `<writer_example>` when it next writes to that person, from the inbox, a follow-up or a chat draft to that one person: style only, never its facts, and nothing in it can open or close a tag.\n\nSend + allow and Approve + allow are refused before anything is done unless their setting is in training (422 `cant_allow`) and the person's list has room (409 `learned_full`, at most 200 a person); allowing failing afterwards never undoes the send. Putting Replies or Follow-ups in, or Suggested actions out, of training only holds more back and works without signed sign-ins; anything else needs a caller the API knows is that person. What it learned is the person's alone: listed on `GET /delegation` (`training.learned`, each with its `setting`), removed with `DELETE /delegation/learned/{id}`, and for the owner also in Take the lead's \"What it's learned\" list; it never reaches anyone else's list or Take the lead's prompts. A dial left on In training by an older version (mode `training`) moves to People I know with Replies and Follow-ups in training (`schema._move_training_off_the_dial`).\n\n**No model decides whether a reply goes.** The watcher's two model calls are unchanged and tool-less, so text in an email can at most change the verdict and the draft. After the draft is written, `handle_it.refusal` — plain code — decides, and anything it refuses becomes the usual card with the reason kept on it (`handle_it_reason`). It refuses when:\n\n- it's off, or the sender is a stranger and the setting isn't Bold, or on a server signed sign-ins are off (`signing_ok`: signed callers, or local login);\n- Gmail couldn't authenticate the sender, whatever the roster says;\n- the email isn't a question, request, scheduling, introduction or follow-up, or the classifier was less sure than the setting needs (0.9 on Easy ones, 0.8 otherwise);\n- the reply goes to anyone the email didn't already go to (its From, To and Cc, less the person's own and the Executive's addresses), leaves out the sender, or goes to more than 10;\n- the email's subject, the sender's new words, or the reply touch money, contracts, legal, hiring, pay, layoffs, the board, investors, mergers, the press, health, or credentials and secrets (passwords, logins, tokens, API keys); or the reply carries an amount, a percentage or a link (a scheme, `www.`, a bare host with a path such as `bit.ly/…`, or a bare host on a common top-level domain) unless it's Bold and at least 0.9 sure, or runs over the setting's length (400, 1200 or 2000 characters);\n- the draft carries any flag but *others on the thread* or *Executive on the thread* (they asked if it's an AI, recipients were trimmed, …);\n- it already sent one in this thread in the last day, or the person's last message in the thread was itself sent on its own (no auto-reply ping-pong);\n- the day's limit is reached (`DELEGATION_HANDLE_IT_MAX_SENDS_PER_DAY`, default 20, within the watcher's drafts limit), or it can't be counted. Any error in the rules counts as a refusal.\n\nA reply the rules allow gets a card with `gate_mode` `auto_execute`, and the watcher sends it at once through `reply_send.send_on_its_own`, the second and last door into the send path (an AST test pins both callers: the approve route and the watcher). It runs the same send-time checks as a tap — switches, mailbox, draft, recipients, the claim and the last re-read — with no caller, and adds its own: the card must be `auto_execute`, Handle it for me still on with the setting still allowing that sender (or, for a follow-up, that recipient), signing still on, the once-a-thread-a-day rule and the daily limit counted again, and the draft untouched since it was saved (an edit in Gmail means the person has taken it over: `draft_changed`, the card waits for them). Every refusal hands the card back to the person (`decision_ledger.hand_back`, a compare-and-set to `propose`), with the reason kept on it when there are words for it, and a failure never stops the scan. Every card it left for the person says why on Today (`waited_because` on `GET /delegation/replies`). A reply sent on its own finishes as `executed` (not approved), is recorded in `delegation_handled` (for the per-thread and daily rules) and writes a private `delegation_reply_handled` row; an unconfirmed one is settled by the reconciler the same way. No History note is taken from it, since notes come only from the person's own words. It runs inside the watcher's scan, so it stops while the Executive is paused.\n\n**Follow-ups as you** (`delegation.follow_ups`). While Handle it for me is on, the watcher's scan also looks at the person's own sent mail, at most once an hour (`LOOK_EVERY`), for an email of theirs that asked something (a question mark in their own words), is 3 to 10 days old, went to at most 10 people other than the Executive, and is still the last message in its thread. Plain code picks it; each gets one follow-up ever (its `delegation_inbox_messages` row), at most two a look, within the drafts limit. The ghostwriter, tool-less as always, writes a short nudge in their voice from the thread, saved as a draft in that thread to exactly the people the email went to. `handle_it.follow_up_refusal` — plain code — decides whether it goes: by the setting it never goes on Easy ones, goes to their team and contacts on People I know and to anyone they wrote to on Most mail (`follow_up_refused` otherwise: `follow_up_level`); it also refuses on any change of recipients, a sensitive topic, an amount or a link, more than 600 characters, a flag, and the same per-thread and daily counts as replies. It goes through `send_on_its_own` on an ordinary `delegation_reply` card marked `source: follow_up`, which rechecks the setting for that recipient. A refused one waits on Today as **Follow-up waiting**, and closes by itself (its untouched draft deleted) when anyone answers in the thread first.\n\n**What it sent** is the person's alone: `GET /delegation/handled` lists the last 7 days (who, subject, the reply, the questions it left them and a link to the thread in their mailbox), and the Settings card shows it under the switch with how many went today.\n\n### Take the lead as you\n\nOn Settings → Act as me, Handle it for me is one dial: **Off**, **Easy ones**, **People I know**, **Most mail** and, for the owner, **Everything**, which uses **Take the lead** (`PUT /delegation/take-the-lead`, one row `person:<id>` in `take_the_lead`; turning it on turns Handle it for me on). While it is on, `handle_it.refusal` drops the setting's limits: strangers, the classifier's confidence, amounts, and the length up to a 4000-character ceiling. A reply with a link still waits for a tap (refusal `link`). In their place it checks the company's and the person's own added rules (`take_the_lead.reply_hit`, refusal `lead_rule`). Everything Handle it always keeps still holds: signed sign-ins, an authenticated sender, a reply kind, exactly the thread's people, the topics that always wait, draft flags, never twice in a row, one a thread a day and the daily limit. Follow-ups under it may go to anyone the person wrote to, with the same sensitive, amount, link and length checks plus the added rules. `reply_send.send_on_its_own` accepts the card while both switches are on. Each person adds and removes their own rules (`/delegation/take-the-lead/rules`). Removing one needs a caller the API can tie to them.\n\n## The mailbox client\n\n`delegation.gmail.DelegateGmail` talks to the Gmail REST API directly and is **never registered with the MCP gateway**, so the model reaches the mailbox only through the typed handler. The credential is one file per person in `DELEGATION_GOOGLE_CREDENTIALS_DIR`, named from a hash of the address (0600 in a 0700 directory), minted by `scripts/connect-own-gmail.py` on the Executive's own OAuth client with the scopes `gmail.readonly` and `gmail.compose`. It never lives in `WORKSPACE_MCP_CREDENTIALS_DIR`, where workspace-mcp picks a credential by address. Besides reading threads and saving drafts it lists the inbox and checks whether they have written to an address, reads one message or draft, lists the person's send-as addresses and deletes a draft. Its **one send method**, `send_draft`, sends an existing draft by its id and nothing else — never `messages.send`, which sends any text it is given — and unit tests pin the method set, the endpoint, the request body and its single caller. A failed request says whether it may have been carried out anyway (a timeout after connecting, a 5xx), which is what the send path needs to know. A parsed message carries when it arrived, whether it is bulk, a bounce or a calendar invite, and whether Gmail's authentication results vouch for its sender.\n\n`gmail_status` answers `connected`, `not_configured`, `needs_reconnect` (Google refused the saved sign-in — an app in Testing mode has its tokens expire after 7 days; a rate limit, though Gmail may send it as a 403, is an `error` to retry, not a reconnect), `mismatch`, `no_email`, `shared_mailbox` or `error`. `shared_mailbox` means the person's address *is* the Executive's; it is refused before asking Google, because then everything the Executive writes would look like theirs. The gateway closes the other door: every `google_workspace__*` call must act as the Executive's own account, and a `user_google_email` naming anyone else is refused and audited (see Integrations).\n\n### Outlook\n\nA person may connect an Outlook mailbox instead (work, school or personal), with `scripts/connect-own-outlook.py`: a device-code sign-in on the Executive's own Entra app (`MS365_MCP_CLIENT_ID`, tenant `MS365_MCP_TENANT_ID` or `common`) asking for `User.Read`, `Mail.ReadWrite`, `Mail.Send` and `offline_access`. It writes the same per-person file with `\"provider\": \"microsoft\"`, so a person has one mailbox for Act as me, and `gmail_for` hands back `delegation.outlook.DelegateOutlook`, which keeps `DelegateGmail`'s methods, return types and errors, so every caller above runs unchanged. It talks to Microsoft Graph directly, never through the Microsoft 365 MCP server the Executive's own Outlook uses, and is never registered with the gateway. Microsoft rotates the refresh token on use; the new one is written back into the file (atomically, keeping every other key, and only if the file still holds the old one). Graph is asked for immutable ids, so a sent draft keeps its id; a thread is Outlook's `conversationId`. Gmail's labels are mapped onto what the callers read: `SENT` only for mail in their Sent Items folder (a From line set to their address by someone else never counts), `DRAFT`, `INBOX`, and Focused Inbox's Other read as a tab. A sender counts as authenticated only when the topmost `Authentication-Results` — the one Exchange wrote, above anything the sender added — holds exactly one DMARC verdict, a pass for From's own domain, and Exchange Online Protection's own `compauth=` verdict, so a header the sender wrote on mail EOP never stamped doesn't count. `delete_draft` refuses anything that isn't a draft, and deletes with the draft's etag (`If-Match`), so a draft sent in between — it keeps its id — is left alone. No client secret is ever written into a person's file, and Graph has no signature setting to read.\n\n## Privacy\n\nWhat a person's mailbox touches is theirs. For the principal that means private to the principal, as every private row is. For a team member it means theirs alone, the principal included: an audit row carries `private_to_person` (set by `audit.rows_for_person` around the inbox watcher, a card's Send and Dismiss and their settings changes, and by `people_tools.audit_row_owner` on a turn that read their mail) and is left out of every audit read — the log, a session's timeline, a row by id, the brief — with only the usage totals counting it; their reply cards are theirs alone (above); and the owner's view of their use is counts only (drafts and sends in 30 days, and whether their inbox watcher is on). The conversation itself becomes its owner's alone before the mailbox is first read: `ghostwrite_email` marks the session `mail_private` (`session_store.mark_mail_private`, creating a channel adapter's row early, with its owner) and refuses to read the mailbox if that can't be stored or the conversation already belongs to someone else (which it then leaves unmarked); from then on only its owner may open it (`chat._session_access`), and working-style learning skips it. Every later turn there starts private (`settings.pin_turn_delegation` reads the flag and fails closed, setting `touched_mail`), so a follow-up answered from history alone is still private to its owner and teaches no memory, and the web chat's rows before the turn (the inbound message and the knowledge lookup) are theirs too. It does not start locked down: the lockdown is `read_mail`, set only by a turn that reads the mailbox or notes itself, and it is never carried into the next turn, because history keeps only the speaker's words and the Executive's replies, never the mail a tool returned. Their next message is them asking again, having read that reply, so everything runs again there. Anyone else speaking in it is not offered Act as me.\n\nOnce `touched_mail` is set, every audit row the turn writes is private (`people_tools.audit_row_private_to_principal`), and a team member's is theirs alone: the tool row (its input and result redacted, since `ghostwrite` is a sensitive tool name), the reply's `chat_turn` row that carries the preview, the model-call rows, and the background passes, which run under `private_rows`. The Honcho peer and department syncs are skipped, because peer memory is read on other people's turns. No memory is learned from the turn at all: the episodic, open-loop and working-style passes are skipped too, since its reply quotes a draft built from other people's mail. The `delegation_*` audit events are always private and carry metadata only — ids, counts and flags, never the draft or the thread. The inbox watcher is private the same way: every row a scan or a card action writes is its person's alone, and it starts no chat turn, alert, scheduled action or workflow run and teaches no memory.\n\nThe turn also keeps one door shut (`delegation.lockdown`). Anyone can put text in the inbox, and that text could try to steer the model, so from the round that calls `ghostwrite_email` or one of the mailbox reads — a round's tools run together — until the turn ends, the dispatch guard refuses only what opens the open internet or reaches everyone: outside fetches (`read_document`, research, the watchlist, `load_mcp_server`, and any MCP `call_tool` outside the private-turn set of reads and recipient-checked sends), `run_script` and `run_python_job`, workflows (they run later, unattended, and may fetch or script), broadcasts, department messages and alerts, and changes to who is on the roster or holds authority (`archive_person`, `set_department_head`, `resolve_roster_request`), and text that steers later turns (`schedule_followup`, whose intent runs unattended when due, and goals and decision outcomes, shown to every later turn). Everything else the person asks for runs, because it is fenced on its own: DMs and `message_person` reach only the roster, invites name attendees by person id, the Executive's own email checks every recipient against the roster, standing facts and profile edits keep only the person's own words, and a workflow drafted in chat is saved only once a person confirms it. `upsert_person` runs only for a contact the person asked for (`speaker_named_contact`: a contact, never the team, an approval scope or a chat account, with a word of the name in what they typed, in a message that asks to add, save or change someone, a new contact's address one they typed or one that sent the mail the turn read under the contact's name (`TurnDelegation.mail_senders`), never one the mail's text only mentions, and on an update a changed address they typed), since the roster is what the send checks trust. Any other contact add is refused with a pointer to an approval card (`lockdown.CONTACT_REFUSAL`), so the Executive leaves an `add_contact` card in the chat rather than asking them to type the address. What an email itself asks for is still never done on its say-so; it goes on an approval card. Every tool is classified as allowed or withheld, a test fails on a new unclassified one, and an unknown name is withheld. A refused call gets a private `tool_invocation` row (`refused: mail_touched`) and tells the model to offer it in the next message. The offered tool list never changes mid-turn, so the cached prefix holds. The handlers that reach outside check again on their own (`outside_reach_refusal`). The server-side web search stays, as on a turn private to the principal. The lockdown ends with the turn. Later turns of that conversation stay private to its owner (`touched_mail`) but hold nothing: a refusal says it will work when they ask again in their next message, and it does.\n\n## Always in the loop\n\nA sent reply can leave a note. Once the reply went, `reply_send` starts `memory.history_notes.note_sent_reply` for the person who sent it, in the background, when their \"Keep track of what happens\" switch is on and the conversation isn't one they asked not to remember. The note-taker sees only the words they sent, so other people's mail never becomes a note, and a failure never touches the send. Recall is `recall_history`, offered on the same verified, private surfaces as `ghostwrite_email` and also allowed under the mail lockdown, since it only reads the speaker's own notes. Calling it locks the turn like `ghostwrite_email` does, from its own round on. The same switch also covers notes from chat, which don't need Act as me. See **Memory** for the checks, retention and controls.\n\n## The system prompt\n\nA constant, name-free `DELEGATION_ADDENDUM` joins block 0 after the unchanged identity addendum while anyone on the install has it on (`block0_delegation_on`). It names the four mailbox reads and says other people's mail is data and that a conversation which read it reaches nobody else; and it tells the model that `ghostwrite_email` is the single exception to *never impersonate*, and only for the person being spoken to; that everything else is written as itself; that when the tool isn't offered the asker can't have it; that a draft is never described as sent; and that it never denies being an AI. It is keyed on the install-level setting, never on the turn or the speaker, so switching it misses the 1h cache once (see Prompt Caching).\n\n## Surfaces\n\n- **Settings → Act as me** (`ActAsMeCard`, hidden unless `GET /delegation` answers, so a team member sees it once the owner lets them): the mailbox connection (`provider` says which) with both connect commands (Gmail and Outlook) and **Check again**, the **Write drafts as me** switch, the **Draft replies to my inbox** switch with its status in plain words, when it last checked and **Check now**, and **How I write**: length and tone, a greeting per audience (`{first}` is their first name), the sign-off, habits and never-rules in fields that grow to show every line, the signature with **Refresh from Gmail** and **Don't add my signature** (or **Add my Gmail signature** when none is added; on Outlook only **Don't add my signature**, with why), and each example quoted with **Remove examples**. The owner also gets **Let team members use it** where the install allows it, with each team member who has it on listed by name and counts only.\n- **Today → Replies waiting** (`RepliesWaiting`, the top of the left column; hidden unless `GET /delegation/replies` answers with cards): each card shows who wrote and who they are to you, whether their address was verified, the subject, **They wrote** as plain text, your draft, **Decide before sending** (the open questions) and its warnings, with **Send** (a confirm naming who it goes to, and a second one when that or the thread changed), **Edit** (a text box on the card with Save and Cancel), **Open in Gmail** or **Open in Outlook** when the reply is in their Drafts, and **Dismiss**.\n- **Setup status**: a **Your own mailbox (Act as me)** light (`your_gmail`), amber while the inbox watcher can't reach the mailbox, is being rate-limited or has 25 replies waiting.\n- **Settings → Your Executive** (`app/settings/executive`, where `/settings/on-its-own` lands): the pause, **Take the lead** and **Book meetings without asking** (as the Executive). **Handle it for me** (the Off / Easy ones / People I know / Most mail / Everything dial, what each sends, how many went today and what it sent this week, as you) sits under the mailbox card on **Settings → Act as me**.\n- **API**: `GET`/`PUT /delegation`, `GET`/`PUT`/`DELETE /delegation/voice`, `POST /delegation/voice/learn`, `POST /delegation/voice/describe`, `POST /delegation/voice/signature`, `PUT /delegation/inbox` (`{enabled?, mailbox_drafts?}`), `POST /delegation/inbox/check`, `GET /delegation/replies`, `PUT /delegation/replies/{id}/text`, `PUT /delegation/handle-it`, `PUT /delegation/training`, `DELETE /delegation/learned/{id}` and `GET /delegation/handled` — the caller's own delegation only — and `PUT /delegation/team`, the owner's switch (see API Reference). Send is `POST /decisions/{id}/approve`; Dismiss is `POST /decisions/{id}/reject`.\n- **Evals**: a scenario's `delegation:` block runs the turn against an in-memory mailbox (`evals.mailbox.ScenarioMailbox`) and hands the saved drafts to the judge (`delegation_001`–`003`). A `type: inbox` scenario runs the watcher's two model calls on one thread: whether it drafts must match the scenario, and a draft is judged on saying nothing the person hadn't, copying no link or address, and listing every open question (`delegation_inbox_001`–`007`; the last checks a draft never claims anything was sent).\n\nActing within rules the person sets and standing in for them are later, separate changes, each behind its own switch, off by default.",
+ "markdown": "The **delegation** module (`openexecutive.delegation`) is **Act as me**: the one place the Executive writes *as* a person instead of as itself. Everywhere else it has its own Google account and bots and signs as itself. Here — when the person asks, or, with a second switch, for mail in their inbox that needs them — it writes an email in their voice and saves it as a **draft in their own mailbox** — Gmail or Outlook — for them to review and send. The one thing it ever sends is one of those inbox drafts, exactly as it is in their mailbox: when the person taps **Send** on its card, or, with a third switch, **Handle it for me**, on its own when plain-code rules allow it (below); chat never sends. The principal can turn it on for themselves, and so can each team member once the owner lets them. It is off by default, and off changes nothing: the prompts, tool lists and prompt caches are byte-identical to an install without it.\n\n## Who can have it, and when it is offered\n\n`delegation.settings.can_delegate(person)` is the single rule for who may turn it on: the principal (a non-archived team member flagged `is_principal`), and any other non-archived team member — never a contact — once team members may have it (`team_members_enabled`). That takes two things: the install allows it (`DELEGATION_TEAM_MEMBERS`, default off, which offers the owner the switch) and the owner turned on **Let team members use it** (`PUT /delegation/team`, principal only, one row in `delegation_team`). Turning the owner's switch off takes it away from every team member from their next message, and their inbox watchers stop at their next check; their own switches, drafts and cards are kept. Everything is keyed by `person_id`. A person turns it on for themselves (`PUT /delegation`); nobody turns it on for anyone else. The setting is one row per person in `delegation_settings`, and a missing, unreadable or zero row all mean off.\n\nAt the start of every chat turn (`Executive.stream_chat` and the committee path) `pin_turn_delegation(session, speaker_text)` runs next to the workspace-mode pin and records a `TurnDelegation` on the session. `ghostwrite_email` is **offered** only when all of these hold:\n\n- the speaker may have it and turned it on — read fresh every turn, so turning it off applies from the next message;\n- the speaker is on a surface that verified it is them — the principal through `people_tools.is_principal_on_verified_surface`, a team member through `teammate_on_verified_surface` (the web app signed in, their own Slack or Discord, never an email turn) — and a web turn also carried a signed-in caller (`Session.web_caller_signed_in`, set from `x-caller-email`) or runs under local login — a request with no caller header is not a sign-in and never gets it (whoever holds the shared secret is trusted as the UI proxy that stamps that header, as on every principal-only route);\n- the conversation is private to the speaker: the web chat, a Slack or Discord DM, or a verified private Telegram chat — never a shared channel or thread, where the draft's preview or the matching threads would be posted for everyone and other people's messages sit in the model's context;\n- the turn is neither unattended (the scheduler's proactive run) nor private to the principal (the email poller's turns), so inbound text can never reach it. The MCP server, workflows, reflection and research never carry it at all.\n\nThe mailbox connection is deliberately **not** part of the offer: the handler checks it on every call and says how to fix it, so the cached tool prefix never flips when a token lapses. The pin lives in a context variable for the turn's task as well as on the session, so two turns running at once on one session (a second browser tab) each read their own. It also carries the speaker's own words for the turn (a new email may go to an address they just typed — the turn's message joins the history only once the turn ends), a per-turn draft count, and `touched_mail` (see Privacy). `own_words` reads those words back for every own-words check (addresses here, facts and task assignments elsewhere): it first removes the hydrated `<outbound_reply_context>` backstory by its exact, anchored shape, so a look-alike closing tag quoted inside it cannot pass the rest off as the speaker's, then drops other adapter tag blocks, and returns nothing for a message carrying an attachment.\n\n## The tool: `ghostwrite_email`\n\nIt lives in its own registry (`orchestrator.delegation_tools.DELEGATION_TOOLS` / `DELEGATION_TOOL_HANDLERS`), **never** `_ALL_SKILL_TOOLS`, so no other toolkit can carry it. On a turn it is offered to, the agent loop adds it to the pool *before* the sort and builds that turn's handler map from both registries; on any other turn a call to it is an unknown tool. The handler re-checks the pin and the surface anyway, then:\n\n1. **Caps** — at most 5 drafts per turn, and `DELEGATION_MAX_DRAFTS_PER_DAY` (default 50, at most 1000) per person per UTC day, shared with the inbox watcher (`delegation.caps`). It is counted from `delegation_drafts`, one row of ids per draft saved from either place, with a count kept in the process as a floor in case a row was lost. Each slot is taken before the first await — a round's tool calls run concurrently, so otherwise every call in it would pass the caps before any counted — and handed back if nothing was drafted. If the rows can't be read it refuses rather than guess.\n2. **Marks the turn** — sets `TurnDelegation.touched_mail` *before* its first read of the mailbox.\n3. **Checks the mailbox** — `gmail_status` asks Google or Microsoft whose mailbox the credential opens and compares it with the person's People email, on every call: a roster email can change, and a client slot swaps the roster.\n4. **Finds the thread** — by `thread_id` (checked against the mailbox's own id shape), or a Gmail-style search (`find`; on Outlook, words, `from:`, `to:`, `cc:`, `subject:` and `newer_than:` are kept and other Gmail operators dropped). Several matches return `{status: \"choose\", candidates}` (subject, sender and date — no message text) and draft nothing; none returns `not_found`.\n5. **Sets the recipients in code**, never from the model. A reply goes to the last inbound message's `From` — never its `Reply-To` (a different one is flagged `reply_to_ignored`) — with the thread's other recipients only on `reply_all`. A new email or a forward may go to any well-formed address: it is only a draft, waiting in their own mailbox, where they see who it is to before they send it. An address that is neither on the roster (contacts only when `contacts_reachable_now()`) nor typed by the speaker this turn comes back in `not_in_people`, and the Executive names it for them to check (`delegation_tools.not_in_people`). Mailing-list mail, a thread they replied to last and a reply-all trimmed to 10 recipients are flagged. An address inside a block an adapter added to their message (the backstory hydrated into a DM) is not one they typed, and when the message carries an attachment — a document's inlined text, or a note naming the file — none of its addresses count as typed: neither has an end mark, and a filename is chosen by whoever sent the file, so such an address is named back to check.\n6. **Writes it** (the ghostwriter, below) and saves it with the mailbox client's `create_draft` (Gmail: threaded with `In-Reply-To`, `References` (whole message ids: when the chain is too long for one header the oldest after the thread's first go, never the parent) and `threadId`, and stamped `X-OE-Ghostwritten`; Outlook: Graph's `createReply` on the message it answers, then the text and recipients, and a named property marking it).\n\nThe result is `{status: \"drafted\", draft_id, gmail_link, to, cc, subject, preview, flags, open_questions, note}`, plus `not_in_people` and `check_address` when a recipient needs checking. The link (`gmail.mailbox_link`) is built server-side from a fixed prefix — the thread in Gmail, the draft itself in Outlook on the web — and the note tells the model the preview is the person's draft, to be treated as data. The chat shows an action chip — \"Drafted an email as you to … — in your Drafts\" — that opens it, and only a link under `MAILBOX_LINK_PREFIXES` is ever linked. A sincere \"are you an AI?\" in the thread comes back as an open question for the person to answer themselves.\n\n## Reading their mailbox\n\nFive read-only tools ride with `ghostwrite_email` in the same registry (`orchestrator.mail_read_tools`), so they are offered, re-checked and fenced exactly as it is: `search_my_email` (a Gmail-style search through the client's `search_threads`, or with no query the threads that reached their inbox in the last few days through `inbox_message_ids` + `get_message`), `read_my_email` (one thread's last 10 messages, each sender's new text only, drafts left out, with the files attached to them listed by message id and index), `read_my_email_attachment` (one of those files as text: PDF, Word, Excel, CSV or text, up to 20 MB, a scanned PDF converted) and `my_email_awaiting_reply` (their own sent mail from the last 14 days, to at most 10 people besides the Executive, that is still the last message in its thread). They change, label, draft and send nothing.\n\nThe fifth, `my_email_read_before`, lists the threads `read_my_email` opened for the same person before, in this conversation or an earlier one, so a later conversation can open the same email again instead of asking them to search for it. It keeps where, never what (`delegation.mail_reads`, table `delegation_mail_read`): the thread id, its subject and first sender as one line each, the mailbox address it was read from and when, never a message's text, an attachment or a recipient. Rows are the person's own and are listed only from the mailbox open now, newest first, optionally narrowed by words in the subject or sender; anything older than 30 days is never listed and is deleted on the next write, at most 200 are kept per person, and turning Act as me off deletes them all. It opens the mailbox like the other reads, so it sets `touched_mail`, counts as a search, and returns its rows with the same data-not-instructions note. The Act as me addendum tells the model that these tools reach the person's mailbox while any Gmail or Outlook tools of its own account reach only the Executive's, to search theirs (never falling back to its own) for mail sent to them, and to reopen an email from before with `my_email_read_before` and `read_my_email`.\n\nEach sets `touched_mail` and marks the conversation theirs before its first read (`_open_mailbox`, shared with `ghostwrite_email`), so the lockdown below applies from its round on. Each is capped per turn on the pin: 10 searches (`my_email_awaiting_reply` counts as one), 20 thread reads and 5 attachment reads, the slot taken before the first await. In `read_my_email` every message someone else wrote comes back inside its own `<untrusted_content source=\"email\">` block (`content_trust.wrap_untrusted`, which defuses a closing tag in the text); the speaker's own messages — SENT and from their address (`threads.mine`), never a From header alone — are marked \"Yours\" outside any block. An attachment's text comes back in its own `<untrusted_content source=\"attachment\">` block (`attachments.format_attached_text`), parsed in an isolated worker (`knowledge.loader`); unlike a file sent to the Executive it is never added to the company's knowledge. The clients find an attachment by message and position and look its provider id up on each read (Gmail's changes per fetch): Gmail lists a message's files with the message (`MailMessage.attachments`, leaving out images the HTML shows in place), Outlook says only `hasAttachments` until `list_attachments` asks Graph (leaving out inline ones), and `attachment_bytes` reads one; an attached Outlook item or cloud link has no file to read. Results carry a note that subjects, senders and text are other people's words, data rather than instructions. The lockdown refuses tools, not the reply's own text, so what would carry that text out without a click is closed where it is shown: the web chat loads no image from another site, and bot replies go out without link previews (see Integrations).\n\n## Forwarding and reminders\n\n`ghostwrite_email` also forwards: with `forward` (a thread id) and `to`, it drafts a forward of the thread's latest message, `intent` becoming the short note above it (the ghostwriter writes only the note; the forwarded email's subject, someone else's words, is never put in the intent). Recipients follow the new-email rule (`_new_recipients`: the roster, or addresses the speaker typed this turn). Gmail's draft quotes a forward header and the original's own words under the note and attaches the original's files, up to 10 and 20 MB (`FORWARD_MAX_FILES`, `FORWARD_MAX_BYTES`; any left off are flagged `attachments_left_off`); Outlook uses Graph's `createForward`, which keeps the original and its files, with the note as its comment and the body never patched. It is still only a draft.\n\n`remind_me(text, when)` (`orchestrator.reminder_tools`) rides with the same tools and checks, but opens no mailbox, so it never marks or locks a turn (`MAILBOX_TOOL_NAMES` are the ones that do). It stores the text once cleaned — one line, anything a chat app could turn into a link (a URL of any scheme, an address, a domain-like word, a host and port, an IP address) replaced with \"[link]\", at most 300 characters — with its time (stored in UTC) in `delegation_reminders` (5 a turn, 25 waiting, a year ahead, refused when the person has no channel). When it is due the scheduler sends that text to the person alone, as \"Reminder you set on <date>: …\", running nothing (see Scheduler); the Slack and Discord DMs it may go by turn link previews off. The chat chip shows the text as it will be sent. A reminder that comes due in the person's quiet hours or leave (when `OUTBOUND_RESPECT_QUIET_HOURS` is on) waits unclaimed and goes once they end, rather than being suppressed and lost. It is allowed after mail is read and in later turns of that conversation: it reaches only the speaker, as plain text. `schedule_followup`, which runs a turn with tools and is listed to everyone, stays refused there.\n\n## Approval cards\n\nA turn that read someone's mail may not message anyone, invite anyone or change the roster (see the lockdown above), yet what the mail asks for is often right. `propose_actions` (`orchestrator.action_card_tools`) leaves the speaker a card instead: up to five actions, each spelled out exactly as it will happen, done only when they tap Approve. It rides with the other Act as me tools and checks (`_writer`), opens no mailbox, and is allowed after mail is read because it does nothing on the turn.\n\n- **Kinds.** `invite` (a meeting with people on the roster), `message` (a direct message to someone on the roster, as written) and `add_contact` (a new contact; the principal's cards only, since contacts are theirs). A team member's card can't name a contact. Each action is checked when the card is made (`delegation.action_cards.check`) and again when it is carried out, so an archived person or a time that has passed fails that action alone.\n- **Whose.** A card is a `delegation_actions` decision for the speaker, `approver_only` like a reply card: only they see it (`GET /delegation/actions`) or resolve it, a team member's answers 404 to the principal too, and it is never an alert. The same actions still waiting are that card, not a new one. 3 cards a turn, 25 waiting, dropped after 7 days.\n- **Approve.** In the chat, under the reply that left it (the `propose_actions` chip carries the card's `decision_id`, and `ChatActionCards` shows the card while it waits), and on the Briefing's Needs you list, with a box per action. Approve (`POST /decisions/{id}/approve`, `edits.only`) needs the card's person signed in (`verified.caller_refusal`) with Act as me on; the card is claimed first so it runs once, then each ticked action runs as stored with no model call, in an unattended session (the principal's with contact egress) under `rows_for_person`, so every audit row it writes is theirs and nothing reaches the team's activity rail: `message_person` (a direct message or nothing; never its Briefing-alert fallback, which more people read), `create_calendar_event`, or a contact row. Approving also needs the person still allowed Act as me (`can_delegate`) and switched on. An invite is booked straight away when the person approving is the one the meeting gate would ask (`calendar_tools.approved_by`), and otherwise goes to that approver as usual, so a card never widens anyone's authority. Per-action results (`done`, `waiting`, `failed`, `skipped`) are kept on the decision and shown on the card. Dismiss rejects it and does nothing else. With Suggested actions in training the card adds Approve + allow, and a card of actions all allowed is carried out on its own (see In training above).\n- **Channels.** A card can come from any surface Act as me runs on; it is approved in the web app, since the chat apps have no buttons here.\n\n## The ghostwriter\n\n`delegation.ghostwriter.compose` writes the text. The Executive decides *what* to say (`intent`); this call decides *how* the person would say it, and it is deliberately small:\n\n- a constant `GHOSTWRITER_PROMPT` plus the person's voice block as the system prompt;\n- the thread (each message cut to its sender's own words, every line scrubbed, all labelled alike), then the writer's own earlier words in it (`<writer_said>`, from their sent mail only), the intent and the recipients' names in the user turn — no company profile, memory or roster;\n- with Always in the loop on, `<writer_noted>`: what the writer's own notes say they told these same people in earlier Act as me replies (`memory.history_drafts.notes_for_draft`: only notes from mail **every** recipient of the new draft was on, so nothing told to one person reaches another; chat notes name no recipient and never qualify; at most 8, newest first, the person's own words or their correction, with dates). The prompt's rule 6 makes it a consistency check, never content: a draft that contradicts it is still written as the intent says, and the conflict becomes an open question on the card, which only the writer sees. Both the chat tool and the inbox watcher pass it;\n- **no tools** but the forced `compose_reply` return, so text in the thread can ask for anything and there is nothing to do it with.\n\nIt is not a Council agent — the Council's edit routes have no principal check — and runs on `DELEGATION_COMPOSER_MODEL` (default: the default model), logged as `ghostwriter` usage. `lint` then runs in code on what comes back: a link or an email address survives only if it appears in the intent, the signature or the recipients, so a link planted in the thread is removed and flagged; the Executive's own name is flagged; and the body is capped, cut at a paragraph, line or word break in its last fifth. A draft with nothing left after that is refused. The signature from the person's Gmail settings (or the one already stored, for Outlook) is appended.\n\n## How I write\n\n`delegation.voice` learns how a person writes *their own* email (`attunement.style` learns how the Executive should write *to* someone). `POST /delegation/voice/learn` reads up to 40 recent sent messages, keeps only what they wrote (quoted replies stripped with the email poller's `sender_new_text`), skips automatic mail, every thread chat drafted into, and every sent message that began as a draft written for them (`delegation_drafts.sent_message_id`), and makes one forced-tool call (`record_voice_profile`). Code validates every field before it is stored:\n\n- greetings per audience (their team, their contacts, anyone else), a sign-off of at most two lines, and a usual length and tone (a backslash-n the model wrote instead of a line break is read as one, keeping the first lines that fit, which also repairs a profile stored before that rule);\n- up to 8 habits and 6 things to avoid, through the same deny-list as working-style rules — no actions, links, handles, amounts, tool names or names of people;\n- up to 3 short exemplars, each checked to be verbatim from a sample, with addresses and phone numbers masked — and none while a client slot is active, so one client's mail never lands in another's state.\n\nThe signature comes from Gmail's `sendAs` settings, not the model (Outlook gives apps no way to read one, so a learn keeps the stored signature and `POST /delegation/voice/signature` answers 409 `signature_unavailable` rather than clearing it), as plain text with its lines kept (`gmail.html_to_text`: a `<br>` always ends a line, and a block tag, opening or closing, ends one when it has text, so Gmail's layout of a first line followed by one `<div>` per line comes out line by line). `POST /delegation/voice/signature` reads it again without relearning: nothing else changes and a locked profile stays locked, so it picks up a changed Gmail signature and undoes **Don't add my signature**. A person can also describe their style in their own words: `POST /delegation/voice/describe` applies the description to the profile they have through the same forced tool (a separate system prompt, no exemplars) and the same validation, keeps their signature and writing examples, and returns the result with a short sample reply for them to review. It saves nothing and needs no mailbox; Save stores it as an ordinary edit. The person can edit the profile, lock it (a locked profile is never relearned) and reset it, and every change is kept in `delegation_voice_history`. One learn runs at a time per person, and a lock or an edit made while it runs wins: the learned profile is dropped (`changed`).\n\n## Draft replies to my inbox\n\nA second switch under Act as me, **Draft replies to my inbox** (`delegation.inbox`), has the Executive watch the person's own inbox and write a first reply to mail that needs them. The reply waits on a private card on Today; nothing is sent until they tap **Send** (below). Where the reply lives is the person's choice, **Also save replies in my Gmail Drafts** (`mailbox_drafts`, off by default): off, it lives on the card alone (`delegation.card_drafts`: the card keeps the draft's recipients, subject and thread headers as `draft_spec`, and Send makes the mailbox draft from exactly that just before the send-time checks, taking it back out if they refuse), so their Drafts folder doesn't fill with replies they may never send; on, it is also saved in their Drafts as before. The switch is one row per person in `delegation_inbox_watch`. Turning it on needs Act as me on and their mailbox connected, and sets `watch_since` to now every time, so it never drafts for mail that arrived before; turning Act as me off turns it off too. Off means no mailbox call and no row written, with one exception: a send the person started that the mailbox never confirmed is still followed through until it is settled.\n\n### When it looks\n\nA throttled hook in the scheduler tick (`_maybe_scan_inbox`, after the pause, company-profile and rotation gates) checks each person whose switch is on every `DELEGATION_INBOX_POLL_MINUTES` (default 5). It is not a scheduled action, because those are listed on `/scheduled` for everyone and anyone may cancel them. One scan task runs at a time, one scan per person, and none while a client slot is active. **Check now** on the Settings card runs the same scan. Every row a scan writes is private and the person's own (`audit.rows_for_person`: a team member's rows are hidden from the principal too, see Privacy), and the process log carries counts and codes, never an address or a line of mail. The scan never marks mail read or changes a label.\n\n### What it drafts for\n\nA scan first settles the open cards (below), and stops with `backlog_full` once 25 are waiting, counting the ones it makes. It then lists the inbox since `watch_since` (`inbox_message_ids`) — not chats, not mail from the person, and not the promotions, social, updates or forums tabs (on Outlook: the Focused inbox only) — up to 100 messages, and takes the newest message of each thread that still needs a look, at most 10: settled threads and threads with an open card are passed over first, so a burst of mail, or a pause for a limit or a back-off, never hides older threads for good. `delegation_inbox_messages` is its durable cursor: one row per message (`UNIQUE(person_id, message_id)`) with what happened to it, the reason, how many tries it took, and the draft and card ids. It stores a hash of the sender's address, never the address. A message is skipped, with the reason recorded, when:\n\n- it came before `watch_since`, or more than three days ago;\n- it is from the person or one of their send-as addresses, from the Executive, or one of this module's own drafts;\n- it is automatic — `Auto-Submitted`, a bulk or auto-reply `Precedence`, a bounce, a calendar invite or a no-reply sender — or from a mailing list, or in a Gmail tab, spam or trash;\n- the person is not in To or Cc, or it has more than 10 recipients;\n- the person replied last, or a draft already sits in the thread.\n\nThree kinds are left unrecorded, so a later scan looks again: mail under 10 minutes old (mail the person answers straight away never gets a card), mail in a thread whose card is still open, and mail a daily limit holds back.\n\n### Deciding, then writing\n\nTwo model calls, and code decides between them. The classifier (`delegation.inbox_classifier`) is one forced-tool call, `classify_email`, returning whether the email needs a personal reply, what kind it is and how sure it is. It sees a few one-lined header lines and the sender's own new words (at most 3000 characters, every line scrubbed) inside one `<email>` block marked as data, and it has no other tools, context or memory, so an email can ask for anything and there is nothing to do it with. A reply is drafted only for a question, a request, scheduling, an introduction or a follow-up, and only when the confidence clears a bar that rises the less the sender is known: 0.6 for their team or a contact (the principal's contacts count only for the principal: a team member's mail is weighed by the team and their own sent mail), 0.7 for someone they have written to before (`has_written_to`, a search of their sent mail), 0.85 for a stranger. Pitches, notes of thanks, FYIs, notifications, newsletters and phishing get nothing, and so does any failure. An email that also went to other people is drafted for only when the person is in To and it asks them themselves: the classifier is told their name, whether they are in To or Cc and how many others it went to, and lists the names the email greets or asks by name (`asked_names`). Code matches those against the person's name, whole name parts only (\"Alexis\" never counts for \"Alex\", and a nickname misses, which means no draft), so mail that greets someone else (\"Good morning, Brennan, could you sign ...\") is never theirs, whatever the model makes of it as a whole. When it names nobody, the model's `asked_of_them` decides, false for a question to the group at large. One that only copies them is never drafted for. Either is recorded `not_needed` with the reason `asks_someone_else`. It runs on `DELEGATION_CLASSIFIER_MODEL`, else the routing model. A sender whose address Gmail couldn't authenticate is handled as a stranger whatever the roster says — anyone can put a colleague's or a client's address in From — so they get the stranger's bar, a holding reply and the stranger limits; the card still says who they claim to be, and that the address isn't verified.\n\nThe reply is written by the same ghostwriter, in the person's voice, but from a fixed intent the module builds rather than words the person typed. Because the intent lets it restate the person's own earlier words, which words are theirs is never written inside the thread, where a sender could imitate any marker with a look-alike: every message there is labelled alike, and the person's own words come in a separate `<writer_said>` block after the thread, taken only from mail their own mailbox sent. Nothing inside the thread can close the thread block or open another. The intent: acknowledge the sender; say again only what the person's *own* earlier messages in the thread already said; commit to nothing new — no yes or no, date, time, price, fact or promise — and say they will get back to them; and put every question it could not answer from their words in **open questions**. A stranger always gets a short holding reply. The lint still strips any link or address that isn't the person's own. The reply goes to the sender (never `Reply-To`) and, on a group email, to everyone else it went to as Cc (reply to all, never the person's own addresses or the Executive); a stranger's holding reply goes to the sender alone. The card says when others were on the email, and its `draft_to` lists everyone, which Send checks.\n\nThe order: record the message, classify, reserve a slot in the shared daily limit, write, save the draft in Gmail (from the alias the mail went to, when it went to one) and store its id, then create the card (idempotent on the message) and mark it drafted. Nothing that fails along the way loses the message. A failure that may pass — the classifier or the writer failing, Gmail refusing the draft, a card that couldn't be made (its draft is deleted first, so no draft is left without a card), any other error with that one message — leaves it for a later scan (`retry`), and it is given up on only after three tries; one message's trouble never holds up the rest. A daily limit reached while it was being read leaves it for later without counting as a try.\n\n### The reply cards\n\nA card is a `delegation_reply` decision whose class is **its approver's alone** (`DecisionClassSpec.approver_only`): only the person it was drafted for sees or resolves it on any decisions route — a team member's card answers 404 to the principal too — and it is never an alert: alerts feed chat turns, which would put other people's mail into memory. It carries the sender with who they are to the person and whether Gmail could verify the address (SPF/DKIM/DMARC, through `fact_confirmation.headers_authenticated`), the subject, what they wrote, the draft, the open questions and its warnings. `GET /delegation/replies` lists the open ones from the database alone, so Today never waits on Google.\n\n- **Send** approves the card, which sends its draft exactly as it is in their mailbox (see *Send on your tap*). **Edit** changes the reply's words right on the card (`PUT /delegation/replies/{id}/text`, `card_drafts.edit`): only its own person, from a caller the API can tie to them (as for Send), never while it is being sent on its own; a reply in their Drafts is rewritten there too (`update_draft_text`, every header kept), and only while it is still the version the card last saw, so a change made in the mailbox is never overwritten. A card edit counts as edited (`approved_with_edit`, and **Do it like this next time** keeps it). **Open in Gmail** (or **Open in Outlook**), shown only when the reply is in their Drafts, opens it there.\n- **Dismiss** rejects the card; the class's after-reject step then deletes the draft if nobody edited it — its message id is still the one it was saved with, since Gmail gives an edited draft a new one, and the Outlook client makes a draft's message id `<id>.<changeKey>` so an edit there changes it too — and keeps it otherwise. What happened is recorded even when Gmail fails.\n- Each scan first **settles** what Gmail already settled: a draft sent from Gmail, a draft deleted there, a reply the person wrote themselves, a thread that is gone, or a card left for 7 days (its draft stays in Gmail). Each is a compare-and-set close with the reason. A newer message from someone else flags the card instead. A card being sent at that moment is left alone.\n\n### Send on your tap\n\n`delegation.reply_send` is the only thing in Act as me that sends, and all it can send is the draft on a card, by its id: `DelegateGmail.send_draft` posts `{\"id\": …}` to Gmail's `drafts.send`, and `DelegateOutlook.send_draft` posts an empty body to that draft's Graph `send`, and nothing else, so no text, recipient or header is ever passed. It has two doors: `send_approved_reply`, reached only from `POST /decisions/{id}/approve` — a class only the card's own person may resolve — and `send_on_its_own`, reached only from the inbox watcher under Handle it for me (below). AST tests pin both call sites, so no model, chat turn, workflow, scheduled job or MCP call can. A tap is checked like this: At send time it checks everything again:\n\n1. **Who is asking.** The caller is the card's person and the API knows it: signed callers are on and the request carries that person's signed assertion (or the operator's, under local login), or the API runs for local login and answers this computer only. Anything else is a 409 `caller_signing_required` — with only the shared secret, whoever holds it could send as the owner — so on a server Send needs signed sign-ins.\n2. **The switches.** Act as me and Draft replies to my inbox are on, and no client slot is active.\n3. **The mailbox.** Their mailbox is connected.\n4. **The draft.** It still exists, in the same thread, and they haven't replied there themselves since; otherwise the card closes.\n5. **Who it goes to.** To, Cc and Bcc hold between 1 and 10 addresses, never the Executive's own, and it is from one of their own addresses.\n6. **What changed.** Recipients that differ from the card's, or a newer message from someone else, need a second yes: the first tap gets a 409 `confirm` naming the recipients and the reason, and the person approves again with what they were shown.\n\nThen it claims the card (`claim_for_execution`, proposed → executing, a compare-and-set, so a double tap sends once) and reads the draft once more: `drafts.send` sends whatever the draft is when Gmail gets the request, so if it was edited since the checks read it — in another tab, by a delegate — nothing is sent and the card comes back (`draft_changed`). Then it sends, and records the result: `approved_with_edit` when the draft was edited in Gmail, the sent message's id, the draft row's `sent_message_id`, and a private `delegation_reply_sent` row. A refusal that proves nothing went — a 4xx, a lapsed sign-in, a rate limit, a draft already gone — hands the card back. A failure that leaves it unclear — a timeout, a 5xx — leaves it `executing` and says to check the Sent folder, and the next scans ask Gmail: their message in the thread means it was sent; anything else is judged on the second scan that sees it: the draft still there and nothing sent means it was not (Gmail deletes a draft it sends), and the card comes back, flagged; the draft gone and nothing sent means it was deleted. Nothing is ever retried on its own. Sending is the person's own act, so it works while the Executive is paused.\n\n### Limits and failures\n\nAt most 5 drafts a scan; `DELEGATION_INBOX_MAX_DRAFTS_PER_DAY` (default 20) a day, inside the shared `DELEGATION_MAX_DRAFTS_PER_DAY` it shares with chat; 200 classifications, 2 drafts per sender (1 for a stranger) and 10 for strangers in all per day; and never more than 25 open cards. A lapsed Gmail sign-in sets the status and is looked at again every 30 minutes; a rate limit, an error reaching Gmail or a crash backs off from 5 minutes, doubling up to 2 hours. Neither turns the switch off: the Settings card shows the status in plain words.\n\n## Handle it for me\n\nA third switch under Act as me, **Handle it for me** (`delegation.handle_it`), lets the inbox watcher send some of its replies itself instead of leaving them on a card. It is one row per person in `delegation_handle_it` (`enabled`, `mode`); no row means off. One setting says how much goes without them, the way quality does on the Agent Council (`handle_it.RULES`): **Easy ones** (`careful`) sends only replies to people they know (someone on their team, a contact or someone they have written to) of at most 400 characters when the classifier is at least 0.9 sure; **People I know** (`balanced`), where it starts, sends replies to people they know up to 1200 characters at 0.8; **Most mail** (`bold`) also answers strangers, up to 2000 characters, and lets a link or an amount through when the classifier is at least 0.9 sure. Everything else waits on a card as before, so turning it on never drafts less. A table from before the setting gets the `mode` column on its next open (`schema._add_handle_it_mode`): someone who had strangers set to send keeps that as Bold, everyone else gets Balanced. Only the person turns it on (`PUT /delegation/handle-it`), from a session the API knows is theirs: the same rule Send uses (`delegation.verified.caller_refusal`), so with only the shared secret it answers 409 `caller_signing_required`. Turning it on needs Act as me and Draft replies to my inbox on, and turning the inbox watcher off (or Act as me, which turns it off) turns it off too.\n\n**In training** (`delegation.training`). Like someone new, each Act as me job can be put in training on its own, from that job's card on the Act as me page, the way Take the lead has it: **Write drafts as me** and **Handle it for me** each pick Off / In training / On (Handle it's In training is Replies in training, with a *Train follow-ups too* box for Follow-ups), and **Suggested actions** picks Always ask / In training; one *What it's learned* list sits at the bottom, each item named by its job (`PUT /delegation/training`, one row per person and setting in `delegation_training`, absent means not in training). In training it asks first and learns who and how from what the person approves; what it learns is a `take_the_lead_allowed` row (feature `act_as_me`, the person's id), theirs alone.\n\n- **Replies.** Every reply waits on its card (`handle_it_reason` `training`) whatever the dial says, and a reply card adds **Send + allow**: it sends that reply on the person's own tap, exactly as Send does, then allows replies to that sender (key `act_as_me|person:<id>|reply|<address>`). While Handle it for me is on, a reply to an allowed sender goes on its own under the dial's limits (a stranger they allowed included) and every check above (the topics that always wait, links and amounts, new recipients, once a thread a day, the daily limit); it's checked again at send time, so a sender removed since waits. Take the lead as you lifts nothing in training: putting Replies or Follow-ups in training turns it off, and it can't be turned on while they are (409 `in_training`).\n- **Follow-ups.** Every follow-up waits (`follow_up_training`); Send + allow on a follow-up card allows follow-ups to exactly the people it went to (`…|follow_up|<sorted addresses>`), whoever they are, under every other follow-up check.\n- **Suggested actions.** An action card (below) adds **Approve + allow**: it does the actions on the tap, then allows each one that happened (a message to that person, `…|message|person:<id>`; a meeting with exactly those people, the person left out, `…|invite|people:<ids>`). A new contact is never allowed. A card proposed later whose every action is allowed is carried out at once, with no tap (`action_cards.run_on_its_own`, `gate_mode` `auto_execute`, status `executed`), only if plain code finds nothing against it (`on_its_own_refusal`): still in training, Act as me on, a server that ties the switch to the person, nothing sensitive, no link or amount, a message of at most 1200 characters, and at most 10 actions carried out this way a day (checked one card at a time); anything refused is handed back to wait on its card as before. An invite carried out this way is never booked as approved by the person: the meeting gate decides as for any meeting nobody tapped, so a time taken from someone's mail can't land on a calendar unasked. The tool result and chip say what was done. The message text in such an action is model-written from the mail and goes without anyone reading it; the allowance covers who and what kind, not the words, which is why sensitive topics, links, amounts and long messages are held back.\n- **Drafts.** Nothing is sent. When the person changes a reply's draft in their mailbox before tapping Send (plain or + allow) and ticks **Do it like this next time**, what they sent is kept (its new text, quoted replies dropped, at most 1000 characters, `…|style|<address>`), and the ghostwriter gets it in its user turn as `<writer_example>` when it next writes to that person, from the inbox, a follow-up or a chat draft to that one person: style only, never its facts, and nothing in it can open or close a tag.\n\nSend + allow and Approve + allow are refused before anything is done unless their setting is in training (422 `cant_allow`) and the person's list has room (409 `learned_full`, at most 200 a person); allowing failing afterwards never undoes the send. Putting Replies or Follow-ups in, or Suggested actions out, of training only holds more back and works without signed sign-ins; anything else needs a caller the API knows is that person. What it learned is the person's alone: listed on `GET /delegation` (`training.learned`, each with its `setting`), removed with `DELETE /delegation/learned/{id}`, and for the owner also in Take the lead's \"What it's learned\" list; it never reaches anyone else's list or Take the lead's prompts. A dial left on In training by an older version (mode `training`) moves to People I know with Replies and Follow-ups in training (`schema._move_training_off_the_dial`).\n\n**No model decides whether a reply goes.** The watcher's two model calls are unchanged and tool-less, so text in an email can at most change the verdict and the draft. After the draft is written, `handle_it.refusal` — plain code — decides, and anything it refuses becomes the usual card with the reason kept on it (`handle_it_reason`). It refuses when:\n\n- it's off, or the sender is a stranger and the setting isn't Bold, or on a server signed sign-ins are off (`signing_ok`: signed callers, or local login);\n- Gmail couldn't authenticate the sender, whatever the roster says;\n- the email isn't a question, request, scheduling, introduction or follow-up, or the classifier was less sure than the setting needs (0.9 on Easy ones, 0.8 otherwise);\n- the reply goes to anyone the email didn't already go to (its From, To and Cc, less the person's own and the Executive's addresses), leaves out the sender, or goes to more than 10;\n- the email's subject, the sender's new words, or the reply touch money, contracts, legal, hiring, pay, layoffs, the board, investors, mergers, the press, health, or credentials and secrets (passwords, logins, tokens, API keys); or the reply carries an amount, a percentage or a link (a scheme, `www.`, a bare host with a path such as `bit.ly/…`, or a bare host on a common top-level domain) unless it's Bold and at least 0.9 sure, or runs over the setting's length (400, 1200 or 2000 characters);\n- the draft carries any flag but *others on the thread* or *Executive on the thread* (they asked if it's an AI, recipients were trimmed, …);\n- it already sent one in this thread in the last day, or the person's last message in the thread was itself sent on its own (no auto-reply ping-pong);\n- the day's limit is reached (`DELEGATION_HANDLE_IT_MAX_SENDS_PER_DAY`, default 20, within the watcher's drafts limit), or it can't be counted. Any error in the rules counts as a refusal.\n\nA reply the rules allow gets a card with `gate_mode` `auto_execute`, and the watcher sends it at once through `reply_send.send_on_its_own`, the second and last door into the send path (an AST test pins both callers: the approve route and the watcher). It runs the same send-time checks as a tap — switches, mailbox, draft, recipients, the claim and the last re-read — with no caller, and adds its own: the card must be `auto_execute`, Handle it for me still on with the setting still allowing that sender (or, for a follow-up, that recipient), signing still on, the once-a-thread-a-day rule and the daily limit counted again, and the draft untouched since it was saved (an edit in Gmail means the person has taken it over: `draft_changed`, the card waits for them). Every refusal hands the card back to the person (`decision_ledger.hand_back`, a compare-and-set to `propose`), with the reason kept on it when there are words for it, and a failure never stops the scan. Every card it left for the person says why on Today (`waited_because` on `GET /delegation/replies`). A reply sent on its own finishes as `executed` (not approved), is recorded in `delegation_handled` (for the per-thread and daily rules) and writes a private `delegation_reply_handled` row; an unconfirmed one is settled by the reconciler the same way. No History note is taken from it, since notes come only from the person's own words. It runs inside the watcher's scan, so it stops while the Executive is paused.\n\n**Follow-ups as you** (`delegation.follow_ups`). While Handle it for me is on, the watcher's scan also looks at the person's own sent mail, at most once an hour (`LOOK_EVERY`), for an email of theirs that asked something (a question mark in their own words), is 3 to 10 days old, went to at most 10 people other than the Executive, and is still the last message in its thread. Plain code picks it; each gets one follow-up ever (its `delegation_inbox_messages` row), at most two a look, within the drafts limit. The ghostwriter, tool-less as always, writes a short nudge in their voice from the thread, saved as a draft in that thread to exactly the people the email went to. `handle_it.follow_up_refusal` — plain code — decides whether it goes: by the setting it never goes on Easy ones, goes to their team and contacts on People I know and to anyone they wrote to on Most mail (`follow_up_refused` otherwise: `follow_up_level`); it also refuses on any change of recipients, a sensitive topic, an amount or a link, more than 600 characters, a flag, and the same per-thread and daily counts as replies. It goes through `send_on_its_own` on an ordinary `delegation_reply` card marked `source: follow_up`, which rechecks the setting for that recipient. A refused one waits on Today as **Follow-up waiting**, and closes by itself (its untouched draft deleted) when anyone answers in the thread first.\n\n**What it sent** is the person's alone: `GET /delegation/handled` lists the last 7 days (who, subject, the reply, the questions it left them and a link to the thread in their mailbox), and the Settings card shows it under the switch with how many went today.\n\n### Take the lead as you\n\nOn Settings → Act as me, Handle it for me is one dial: **Off**, **Easy ones**, **People I know**, **Most mail** and, for the owner, **Everything**, which uses **Take the lead** (`PUT /delegation/take-the-lead`, one row `person:<id>` in `take_the_lead`; turning it on turns Handle it for me on). While it is on, `handle_it.refusal` drops the setting's limits: strangers, the classifier's confidence, amounts, and the length up to a 4000-character ceiling. A reply with a link still waits for a tap (refusal `link`). In their place it checks the company's and the person's own added rules (`take_the_lead.reply_hit`, refusal `lead_rule`). Everything Handle it always keeps still holds: signed sign-ins, an authenticated sender, a reply kind, exactly the thread's people, the topics that always wait, draft flags, never twice in a row, one a thread a day and the daily limit. Follow-ups under it may go to anyone the person wrote to, with the same sensitive, amount, link and length checks plus the added rules. `reply_send.send_on_its_own` accepts the card while both switches are on. Each person adds and removes their own rules (`/delegation/take-the-lead/rules`). Removing one needs a caller the API can tie to them.\n\n\n## AI line on replies\n\nReplies written as a person can end with **\"Drafted and sent by my AI assistant.\"** (`delegation.disclosure.LINE`), by two switches that are each person's own (`PUT /delegation/disclosure`, one row per person in `delegation_disclosure`, NULL following the install): **Handle it**, on by default (`AI_DISCLOSURE_HANDLE_IT`), for a reply Handle it for me will send on its own, which nobody reads before it goes; and **Approved sends**, off by default (`AI_DISCLOSURE_APPROVED_SENDS`), for one that waits on a card for them to send, since they read it first. The line is written into the reply when it is drafted (`disclosure.sign`, in `inbox._consider` and `follow_ups`), once the plain-code checks have decided whether it goes on its own, so the card, the mailbox draft and what is sent are the same words and Send's draft-changed checks see them; a switch changed later applies to the next reply, and a reply drafted to go on its own that is handed back keeps its line for the person to edit. `AI_DISCLOSURE` off turns both off and hides the card. Turning a switch on only adds the line, so it needs no signed sign-in; turning one off does. Settings → Delegate → Act as me shows the card (**Replies sent from your mailbox**), and turning off the Handle it line asks first.\n\n## The mailbox client\n\n`delegation.gmail.DelegateGmail` talks to the Gmail REST API directly and is **never registered with the MCP gateway**, so the model reaches the mailbox only through the typed handler. The credential is one file per person in `DELEGATION_GOOGLE_CREDENTIALS_DIR`, named from a hash of the address (0600 in a 0700 directory), minted by `scripts/connect-own-gmail.py` on the Executive's own OAuth client with the scopes `gmail.readonly` and `gmail.compose`. It never lives in `WORKSPACE_MCP_CREDENTIALS_DIR`, where workspace-mcp picks a credential by address. Besides reading threads and saving drafts it lists the inbox and checks whether they have written to an address, reads one message or draft, lists the person's send-as addresses and deletes a draft. Its **one send method**, `send_draft`, sends an existing draft by its id and nothing else — never `messages.send`, which sends any text it is given — and unit tests pin the method set, the endpoint, the request body and its single caller. A failed request says whether it may have been carried out anyway (a timeout after connecting, a 5xx), which is what the send path needs to know. A parsed message carries when it arrived, whether it is bulk, a bounce or a calendar invite, and whether Gmail's authentication results vouch for its sender.\n\n`gmail_status` answers `connected`, `not_configured`, `needs_reconnect` (Google refused the saved sign-in — an app in Testing mode has its tokens expire after 7 days; a rate limit, though Gmail may send it as a 403, is an `error` to retry, not a reconnect), `mismatch`, `no_email`, `shared_mailbox` or `error`. `shared_mailbox` means the person's address *is* the Executive's; it is refused before asking Google, because then everything the Executive writes would look like theirs. The gateway closes the other door: every `google_workspace__*` call must act as the Executive's own account, and a `user_google_email` naming anyone else is refused and audited (see Integrations).\n\n### Outlook\n\nA person may connect an Outlook mailbox instead (work, school or personal), with `scripts/connect-own-outlook.py`: a device-code sign-in on the Executive's own Entra app (`MS365_MCP_CLIENT_ID`, tenant `MS365_MCP_TENANT_ID` or `common`) asking for `User.Read`, `Mail.ReadWrite`, `Mail.Send` and `offline_access`. It writes the same per-person file with `\"provider\": \"microsoft\"`, so a person has one mailbox for Act as me, and `gmail_for` hands back `delegation.outlook.DelegateOutlook`, which keeps `DelegateGmail`'s methods, return types and errors, so every caller above runs unchanged. It talks to Microsoft Graph directly, never through the Microsoft 365 MCP server the Executive's own Outlook uses, and is never registered with the gateway. Microsoft rotates the refresh token on use; the new one is written back into the file (atomically, keeping every other key, and only if the file still holds the old one). Graph is asked for immutable ids, so a sent draft keeps its id; a thread is Outlook's `conversationId`. Gmail's labels are mapped onto what the callers read: `SENT` only for mail in their Sent Items folder (a From line set to their address by someone else never counts), `DRAFT`, `INBOX`, and Focused Inbox's Other read as a tab. A sender counts as authenticated only when the topmost `Authentication-Results` — the one Exchange wrote, above anything the sender added — holds exactly one DMARC verdict, a pass for From's own domain, and Exchange Online Protection's own `compauth=` verdict, so a header the sender wrote on mail EOP never stamped doesn't count. `delete_draft` refuses anything that isn't a draft, and deletes with the draft's etag (`If-Match`), so a draft sent in between — it keeps its id — is left alone. No client secret is ever written into a person's file, and Graph has no signature setting to read.\n\n## Privacy\n\nWhat a person's mailbox touches is theirs. For the principal that means private to the principal, as every private row is. For a team member it means theirs alone, the principal included: an audit row carries `private_to_person` (set by `audit.rows_for_person` around the inbox watcher, a card's Send and Dismiss and their settings changes, and by `people_tools.audit_row_owner` on a turn that read their mail) and is left out of every audit read — the log, a session's timeline, a row by id, the brief — with only the usage totals counting it; their reply cards are theirs alone (above); and the owner's view of their use is counts only (drafts and sends in 30 days, and whether their inbox watcher is on). The conversation itself becomes its owner's alone before the mailbox is first read: `ghostwrite_email` marks the session `mail_private` (`session_store.mark_mail_private`, creating a channel adapter's row early, with its owner) and refuses to read the mailbox if that can't be stored or the conversation already belongs to someone else (which it then leaves unmarked); from then on only its owner may open it (`chat._session_access`), and working-style learning skips it. Every later turn there starts private (`settings.pin_turn_delegation` reads the flag and fails closed, setting `touched_mail`), so a follow-up answered from history alone is still private to its owner and teaches no memory, and the web chat's rows before the turn (the inbound message and the knowledge lookup) are theirs too. It does not start locked down: the lockdown is `read_mail`, set only by a turn that reads the mailbox or notes itself, and it is never carried into the next turn, because history keeps only the speaker's words and the Executive's replies, never the mail a tool returned. Their next message is them asking again, having read that reply, so everything runs again there. Anyone else speaking in it is not offered Act as me.\n\nOnce `touched_mail` is set, every audit row the turn writes is private (`people_tools.audit_row_private_to_principal`), and a team member's is theirs alone: the tool row (its input and result redacted, since `ghostwrite` is a sensitive tool name), the reply's `chat_turn` row that carries the preview, the model-call rows, and the background passes, which run under `private_rows`. The Honcho peer and department syncs are skipped, because peer memory is read on other people's turns. No memory is learned from the turn at all: the episodic, open-loop and working-style passes are skipped too, since its reply quotes a draft built from other people's mail. The `delegation_*` audit events are always private and carry metadata only — ids, counts and flags, never the draft or the thread. The inbox watcher is private the same way: every row a scan or a card action writes is its person's alone, and it starts no chat turn, alert, scheduled action or workflow run and teaches no memory.\n\nThe turn also keeps one door shut (`delegation.lockdown`). Anyone can put text in the inbox, and that text could try to steer the model, so from the round that calls `ghostwrite_email` or one of the mailbox reads — a round's tools run together — until the turn ends, the dispatch guard refuses only what opens the open internet or reaches everyone: outside fetches (`read_document`, research, the watchlist, `load_mcp_server`, and any MCP `call_tool` outside the private-turn set of reads and recipient-checked sends), `run_script` and `run_python_job`, workflows (they run later, unattended, and may fetch or script), broadcasts, department messages and alerts, and changes to who is on the roster or holds authority (`archive_person`, `set_department_head`, `resolve_roster_request`), and text that steers later turns (`schedule_followup`, whose intent runs unattended when due, and goals and decision outcomes, shown to every later turn). Everything else the person asks for runs, because it is fenced on its own: DMs and `message_person` reach only the roster, invites name attendees by person id, the Executive's own email checks every recipient against the roster, standing facts and profile edits keep only the person's own words, and a workflow drafted in chat is saved only once a person confirms it. `upsert_person` runs only for a contact the person asked for (`speaker_named_contact`: a contact, never the team, an approval scope or a chat account, with a word of the name in what they typed, in a message that asks to add, save or change someone, a new contact's address one they typed or one that sent the mail the turn read under the contact's name (`TurnDelegation.mail_senders`), never one the mail's text only mentions, and on an update a changed address they typed), since the roster is what the send checks trust. Any other contact add is refused with a pointer to an approval card (`lockdown.CONTACT_REFUSAL`), so the Executive leaves an `add_contact` card in the chat rather than asking them to type the address. What an email itself asks for is still never done on its say-so; it goes on an approval card. Every tool is classified as allowed or withheld, a test fails on a new unclassified one, and an unknown name is withheld. A refused call gets a private `tool_invocation` row (`refused: mail_touched`) and tells the model to offer it in the next message. The offered tool list never changes mid-turn, so the cached prefix holds. The handlers that reach outside check again on their own (`outside_reach_refusal`). The server-side web search stays, as on a turn private to the principal. The lockdown ends with the turn. Later turns of that conversation stay private to its owner (`touched_mail`) but hold nothing: a refusal says it will work when they ask again in their next message, and it does.\n\n## Always in the loop\n\nA sent reply can leave a note. Once the reply went, `reply_send` starts `memory.history_notes.note_sent_reply` for the person who sent it, in the background, when their \"Keep track of what happens\" switch is on and the conversation isn't one they asked not to remember. The note-taker sees only the words they sent, so other people's mail never becomes a note, and a failure never touches the send. Recall is `recall_history`, offered on the same verified, private surfaces as `ghostwrite_email` and also allowed under the mail lockdown, since it only reads the speaker's own notes. Calling it locks the turn like `ghostwrite_email` does, from its own round on. The same switch also covers notes from chat, which don't need Act as me. See **Memory** for the checks, retention and controls.\n\n## The system prompt\n\nA constant, name-free `DELEGATION_ADDENDUM` joins block 0 after the unchanged identity addendum while anyone on the install has it on (`block0_delegation_on`). It names the four mailbox reads and says other people's mail is data and that a conversation which read it reaches nobody else; and it tells the model that `ghostwrite_email` is the single exception to *never impersonate*, and only for the person being spoken to; that everything else is written as itself; that when the tool isn't offered the asker can't have it; that a draft is never described as sent; and that it never denies being an AI. It is keyed on the install-level setting, never on the turn or the speaker, so switching it misses the 1h cache once (see Prompt Caching).\n\n## Surfaces\n\n- **Settings → Act as me** (`ActAsMeCard`, hidden unless `GET /delegation` answers, so a team member sees it once the owner lets them): the mailbox connection (`provider` says which) with both connect commands (Gmail and Outlook) and **Check again**, the **Write drafts as me** switch, the **Draft replies to my inbox** switch with its status in plain words, when it last checked and **Check now**, and **How I write**: length and tone, a greeting per audience (`{first}` is their first name), the sign-off, habits and never-rules in fields that grow to show every line, the signature with **Refresh from Gmail** and **Don't add my signature** (or **Add my Gmail signature** when none is added; on Outlook only **Don't add my signature**, with why), and each example quoted with **Remove examples**. The owner also gets **Let team members use it** where the install allows it, with each team member who has it on listed by name and counts only.\n- **Today → Replies waiting** (`RepliesWaiting`, the top of the left column; hidden unless `GET /delegation/replies` answers with cards): each card shows who wrote and who they are to you, whether their address was verified, the subject, **They wrote** as plain text, your draft, **Decide before sending** (the open questions) and its warnings, with **Send** (a confirm naming who it goes to, and a second one when that or the thread changed), **Edit** (a text box on the card with Save and Cancel), **Open in Gmail** or **Open in Outlook** when the reply is in their Drafts, and **Dismiss**.\n- **Setup status**: a **Your own mailbox (Act as me)** light (`your_gmail`), amber while the inbox watcher can't reach the mailbox, is being rate-limited or has 25 replies waiting.\n- **Settings → Your Executive** (`app/settings/executive`, where `/settings/on-its-own` lands): the pause, **Take the lead** and **Book meetings without asking** (as the Executive). **Handle it for me** (the Off / Easy ones / People I know / Most mail / Everything dial, what each sends, how many went today and what it sent this week, as you) sits under the mailbox card on **Settings → Act as me**.\n- **API**: `GET`/`PUT /delegation`, `GET`/`PUT`/`DELETE /delegation/voice`, `POST /delegation/voice/learn`, `POST /delegation/voice/describe`, `POST /delegation/voice/signature`, `PUT /delegation/inbox` (`{enabled?, mailbox_drafts?}`), `POST /delegation/inbox/check`, `GET /delegation/replies`, `PUT /delegation/replies/{id}/text`, `PUT /delegation/handle-it`, `PUT /delegation/training`, `DELETE /delegation/learned/{id}` and `GET /delegation/handled` — the caller's own delegation only — and `PUT /delegation/team`, the owner's switch (see API Reference). Send is `POST /decisions/{id}/approve`; Dismiss is `POST /decisions/{id}/reject`.\n- **Evals**: a scenario's `delegation:` block runs the turn against an in-memory mailbox (`evals.mailbox.ScenarioMailbox`) and hands the saved drafts to the judge (`delegation_001`–`003`). A `type: inbox` scenario runs the watcher's two model calls on one thread: whether it drafts must match the scenario, and a draft is judged on saying nothing the person hadn't, copying no link or address, and listing every open question (`delegation_inbox_001`–`007`; the last checks a draft never claims anything was sent).\n\nActing within rules the person sets and standing in for them are later, separate changes, each behind its own switch, off by default.",
"mermaid": "flowchart TD\n Ask[\"Principal or team member, verified<br/>'reply to Dana as me: yes'\"] --> Pin[\"pin_turn_delegation<br/>may have it? on? verified? private? attended?\"]\n Pin -->|\"offered\"| Loop[\"Executive tool loop<br/>with ghostwrite_email\"]\n Pin -->|\"not offered\"| Self[\"writes as itself<br/>(unknown tool this turn)\"]\n Loop --> Tool[\"ghostwrite_email handler<br/>caps, mailbox check, recipients in code\"]\n Tool -->|\"touched_mail\"| Private[\"turn's rows and conversation<br/>theirs alone, no memory\"]\n Tool --> Client[\"DelegateGmail or DelegateOutlook<br/>REST, outside the MCP gateway\"]\n Tick[\"Scheduler tick or Check now<br/>Draft replies to my inbox on\"] --> Scan[\"inbox scan, rows theirs alone<br/>skip rules, limits, ledger\"]\n Scan --> Client\n Client --> Box[(\"The person's own Gmail or Outlook\")]\n Scan --> Classify[\"classifier, no tools<br/>needs a reply? kind, confidence\"]\n Classify -->|\"worth a draft\"| Gw\n Voice[(\"delegation_voice<br/>How I write\")] --> Gw[\"ghostwriter<br/>no tools: voice + intent + thread\"]\n Tool --> Gw\n Gw --> Lint[\"lint: links and addresses<br/>only the person's own\"]\n Lint --> Draft[\"create_draft, threaded\"]\n Draft --> Box\n Draft --> Chip[\"chat chip: open the draft<br/>chat never sends\"]\n Draft --> Card[\"Today: Replies waiting<br/>their own card\"]\n Card -->|\"Send, confirmed\"| SendPath[\"reply_send: checks again,<br/>claim, send the draft by id\"]\n SendPath --> Box\n class Ask entry\n class Pin compute\n class Loop compute\n class Self compute\n class Tool compute\n class Private compute\n class Client compute\n class Tick entry\n class Scan compute\n class Classify compute\n class Box external\n class Voice storage\n class Gw compute\n class Lint compute\n class Draft hot\n class Chip entry\n class Card entry\n class SendPath hot",
- "generated_at": "2026-10-10T16:26:53Z"
+ "generated_at": "2026-10-13T15:30:00Z"
}
diff --git a/packages/core/openexecutive/architecture/prebuilt/integrations.json b/packages/core/openexecutive/architecture/prebuilt/integrations.json
index 459f290d..13de61b4 100644
--- a/packages/core/openexecutive/architecture/prebuilt/integrations.json
+++ b/packages/core/openexecutive/architecture/prebuilt/integrations.json
@@ -1,7 +1,7 @@
{
"section_id": "integrations",
"title": "Integrations",
- "markdown": "External channels connect callers to the Executive. Each inbound adapter resolves the sender to a Person — the chat adapters via `find_person_by_{slack,discord,telegram}_id`, the email poller via `people.identity.resolve_email_sender` (a primary address or an alias, then the company-domain rule; see Org) — and threads `person_id` into `Executive.chat()`. A sender who resolves to nobody is no longer simply dropped: see **New senders (roster requests)** below.\n\n### Inbound channels\n\n- **Slack** (`integrations.slack_bot`) — async Bolt app (`AsyncApp`) embedded in the lifespan, hearing Slack one of two ways: socket mode (`AsyncSocketModeHandler`, with `SLACK_APP_TOKEN`) or, with only `SLACK_SIGNING_SECRET`, the Events API over HTTP at `POST /webhook/slack` (`integrations.slack_events`), where the route checks each request's signature and age (and a 1 MB cap) before Bolt sees it, so unsigned traffic never reaches Bolt's logging, and Bolt acknowledges at once while the listener runs. Either way a repeated event id is dropped (Slack resends what it thinks went unacknowledged), as is an event delivered for another workspace. Responds to `@mention`s, DMs, and thread continuations, with per-conversation session memory (see **Conversation state** below). Text only: there is no interactivity endpoint and no Block Kit, so nothing Slack sends carries a button.\n- **Discord** (`integrations.discord_bot`) — discord.py gateway embedded in the lifespan; per-thread auto-titling, session-scoped memory, author attribution, configurable `@mention` thread promotion.\n- **Telegram** — webhook gating + per-chat history. Replies (and every outbound Telegram message) go out as Telegram's HTML subset (`utils.markdown_telegram`: markdown-it-py with raw HTML and images off, text escaped), so the Executive's Markdown shows as formatting; links show as their text and target in plain text and never become tags, so quoted text can't hide where one goes; a chunk Telegram refuses is sent again as plain text.\n- **Google Chat** — webhook + service-account auth. It resolves no approval gates and has no roster gate (the webhook authenticates Google, not the sender), and it never populates the session's trusted alert ids, so `ack_alert`'s trust check — which now runs on every session rather than only those declaring an `origin_channel` — refuses there by default, and its `<channel>` block says plainly that a sign-off cannot be answered from it.\n- **Email** (`integrations.email_poller`) — MCP poll loop over the Executive's own mailbox — Gmail via `workspace-mcp` or Outlook via `ms-365-mcp-server`, chosen by `EMAIL_PROVIDER` (see *Workspace providers* below); it starts only once the MCP gateway is up, never surfaces Reply-To redirects, filters auto-replies. Its sessions carry `origin_channel=\"email\"`: the untrusted-content policy (`orchestrator.content_trust`) reads that tag to tell mail from the principal's own surfaces — left empty, as it once was, every stranger's body reached the decision extractor as the principal's words. The mail goes into the turn inside an `<untrusted_content source=\"email\" author=\"<sender>\" author_verified=\"no\">` block that says it carries none of the principal's authority, with the framing line and any `[POLICY]` notice outside it; mail from the principal's primary address that Gmail authenticated (`dmarc=pass`) is their own words — but only their new text goes in unwrapped: when the mail quotes or forwards another message (`_new_text_lines` cut something), the message as received follows in an `<untrusted_content source=\"email_quoted\">` block, since a stranger's text they replied to or passed on is not theirs. While the Executive is paused (`scheduler.pause`) it skips each poll and leaves the inbox unread, so held mail is processed on the first poll after resume; the chat adapters above are not paused. The Executive sees the whole message; peer memory records only `_email_memory_text` — the subject, the sender's new text (cut at the first quoted or forwarded message) and the attachment filenames — parsed from the rendered message with any trailing `--- REPLY ---` block dropped first, without backtracking-prone regexes, since the text is sender-controlled, and only for a rostered sender, the only kind with a peer to record into. The poller's `integration_inbound` row keeps the mail's text for the audit log, the way the reply's `chat_turn` row keeps the Executive's answer: `full.message` is the sender's new text (what web chat keeps for a turn), `full.attachments` the attachment filenames, and `details.preview` the first 120 characters of the new text, so /audit shows what was asked next to what was answered (the summary still reads `Inbound email from <sender>: <subject>`). The whole body with its quoted chain (`full.body`) is kept only on a row private to the principal: the chain is other people's mail — a contact's, on the principal's reply to one — and every signed-in user reads a public row. Each field is cut by its escaped JSON length (20,000 characters of text, 8,000 of attachment names), the measure the logger's 64 KB cap uses, so a long mail in a non-Latin script loses its tail rather than the row's structure. Outbound, the principal briefs and the client-rotation digest are emailed through the same gateway with the configured backend's send tool (`google_workspace__send_gmail_message` or `microsoft_365__send-mail`, via `workspace.registry.send_from_executive`, from the Executive's own mailbox to the principal's address, subject `<brief title> — <date>`, the date in the user's zone) as formatted HTML — rendered from the brief's Markdown with raw HTML and images off, so text quoted from outside shows as text and opening the email loads nothing remote. Email is first when the principal prefers it, and otherwise the backup after every chat channel, tried only when Slack, Discord and Telegram are missing or fail; an in-band error from the tool counts as a failed send and the next channel is tried. An emailed brief is not a DM, so the per-Person outbound guard below does not apply — the gateway's roster egress gate does.\n\n Two senders get their own framing. Mail from one of the principal's **contacts** (a `kind=\"contact\"` People row — see Org) replaces the non-roster `[POLICY]` notice with a contact one: it names them (name and role/company), says not to reply unless the principal asks and not to tell anyone else, and asks for a summary for the principal instead. Both this turn and the forwarded-by-principal turn below are marked `Session.private_to_principal`: they can email, DM or invite only the principal, an alert they raise is private to the principal, and they are not offered the tools that reach anyone else or start work outside the turn (`schedule_tools.PRIVATE_TURN_WITHHELD_TOOLS`: broadcasts, meeting bookings, workflows, follow-ups, watches, artifacts, goal updates, skill drafts, connecting a new MCP server), and of the MCP tools only Google Workspace reads and the gated Gmail send (see MCP below). Every audit row about such mail is private to the principal (see Audit Log), the poller's two `integration_inbound` rows included (`_private_to_principal_mail`), so the audit log's other readers see none of it. The poller holds `audit.private_rows` for the whole handling of such a mail, because some rows are written before the turn binds its private session: the knowledge retrieval quotes the mail's From line, subject and opening. The rows themselves read exactly like an outside sender's: the `integration_inbound` row is the non-roster one (`accepted_non_roster`), and the notice opens with the non-roster notice's exact sentence. The poller handles each message with no session bound, so one message's private turn never marks the next message's rows. The turn stays unattributed (`person_id` None — no peer memory, no co-presence), and the gateway refuses mail to a contact on a poller turn anyway. Mail **forwarded by the principal** — the sender resolves to the principal and `_new_text_lines` reports a forwarded message — is wrapped in a `<forwarded_by_principal>` block: summarise it, draft a reply the principal could send (as text in the reply, never sent or saved as a Gmail draft), note any commitment or date, and offer to add the original sender as a contact (confirmed from the web app or the principal's Slack / Discord, since an ordinary email turn cannot change the People list — the one email that can is the principal's tokened reply to a roster-request email, which is parsed before any model turn; see below). The reply still goes to the principal only. The From header is unauthenticated, so this framing grants nothing: a spoofed \"forward\" just gets the real principal a summary.\n\nSlack, Discord, and Telegram are roster-gated on inbound: an unrostered sender's message never reaches the Executive (`details.outcome=\"rejected_unknown_sender\"`). The gates look up **team members only** — a contact's Slack, Discord or Telegram id is treated as unknown, so a contact can never talk to the bot (nor count as a co-present peer). What happens next is the roster request below — except for a contact, a thread continuation, a Telegram group or an unverified Telegram webhook, which stay silent. Email is not inbound-gated — non-roster mail flows through with a `[POLICY]` notice so triage can classify it. Slack/Discord thread continuations pass the shared `integrations.response_gate` (utility-fast model, biased toward replying). Slack only continues a thread it has replied in: it reads the thread with `conversations_replies` (5 s bound) and looks for its own user id, which `auth_test` resolves at startup. If that lookup fails, mentions and DMs still work and thread continuations stop until a retry succeeds (`slack_bot._resolve_bot_user_id`, at most once a minute, from the channel `message` listener). When Slack doesn't return the thread in time, the thread's stored history (`slack:thread:{channel}:{thread_ts}`, written on every turn the bot answers there) stands in: with history, a rostered sender gets an ephemeral \"I couldn't read this thread just now… @-mention me\" and the drop is audited (`outcome=dropped_thread_unreadable`). A thread the bot never answered in stays silent. A message the gate skips is audited (`outcome=skipped_gate`, with the gate's reason) and gets an :eyes: reaction (`reactions_add`, which needs the bot's `reactions:write` scope; without it the reaction is skipped), so a sender the gate misjudged can see the message was read and @-mention the bot.\n\n**No link previews on replies.** The Executive's replies go out with previews off — Telegram `sendMessage` carries `link_preview_options.is_disabled`, Slack's reply `say` passes `unfurl_links=False` / `unfurl_media=False`, and every Discord reply send passes `suppress_embeds=True`. The direct messages the Executive starts are the same: `send_slack_dm` posts with `unfurl_links=False` / `unfurl_media=False`, and Discord's `send_dm` sets the `SUPPRESS_EMBEDS` flag on each chunk, so a reminder or a brief sent by DM carries no preview either. Otherwise the chat service fetches a link in the text on its own, so a reply that quotes someone's mail (Act as me reads it) could carry data to a sender's server with no click. The web chat does the same for images: `Message.tsx` loads only an image the app itself serves (`lib/markdownImages.loadsInline`) and shows any other as a link.\n\n### Alert triage of chat messages\n\nEvery chat-app message that passes the roster gate (Telegram, Slack, Discord, Google Chat) is also handed to alert triage through `alerts.pipeline.schedule_chat_message_evaluation`. The sender is named by their People record (`full_name`), not by the name they typed into the chat app, so a person whose account there is called something else is never read as a stranger; the chat app's own name (or raw user id) is used only when nobody on the roster matched. The principal's own messages to their Executive are a conversation, not news for their brief, so they are not triaged at all. Triage sees the sender as `sender:` and the room as `channel:`.\n\n### New senders (roster requests)\n\nSomeone who writes in and is not on the People list — and is not one of the principal's contacts — is **held, acknowledged and asked about** instead of dropped (`integrations.roster_intake.intake` over the `people.roster_requests` ledger; lifecycle in Org). Each adapter calls it after its gate:\n\n- **Slack** — a DM or an `@mention` (not a thread continuation: the bot was not addressed there). The name and profile email come from `users_info` (5 s timeout). The acknowledgement goes in the DM, or as an **ephemeral** message for a mention, so nobody else in the channel learns who is and is not on the roster. The held event keeps only a fixed set of keys (text, user, channel, ts, thread, files).\n- **Discord** — a DM or a mention (not a thread the bot merely follows). The acknowledgement goes in the DM, or by DM for a channel mention; `/ask` answers the interaction with an ephemeral acknowledgement first (Discord allows 3 s), then holds the prompt. Its old \"ask the admin to add you\" reply is gone.\n- **Telegram** — only a **private** chat, and only when `TELEGRAM_WEBHOOK_SECRET` is set and valid: Telegram identifies the chat rather than the sender, and without the secret anyone could forge an update. A group, or an unverified webhook, stays silent as before. Attachment file ids are held and downloaded only on replay.\n- **Email** — non-roster mail that is not private to the principal (so never a contact's) and does not look machine-sent (`looks_automated`: a no-reply / notifications / billing-style local part, a `List-Id` / `List-Unsubscribe` / `Auto-Submitted` / bulk `Precedence` header, or a DMARC failure). The mail is **still triaged** by the Executive on this pass; its `[POLICY]` notice now says the sender was already told their message is waiting and the principal was asked who they are, so the model raises no second alert just to add them, and that the email will come back once they are added.\n\nEvery path skips a contact first (`find_person_by_*_id(..., include_contacts=True)`, or the email poller's private-mail rule): a contact has no chat access and gets no request and no acknowledgement, and the team-visible `rejected_unknown_sender` / `accepted_non_roster` rows read the same for a contact as for a stranger.\n\n**The acknowledgement** is one fixed text (`roster_intake.ACK_TEXT` — \"your message arrived … waiting for confirmation\"), never the model and never an echo of what they wrote, so a forged sender gains nothing by triggering it. It is sent at most once per sender per `ROSTER_ACK_WINDOW_DAYS` (7) and at most `ROSTER_ACK_DAILY_CAP` (20) a day across all senders (`claim_ack`, one transaction; a failed send releases its claim). An email acknowledgement is the one send the MCP gateway lets reach an address off the roster: `send_email_ack` sets `mcp_gateway.roster_ack_grant`, a one-shot ContextVar grant that admits a single `send_gmail_message` with exactly the granted recipient (a plain string, no list), subject and body and no other key, then is spent (audited `integration_outbound`, `kind: roster_ack`, private). An email sender is acknowledged only when Gmail authenticated them: once `claim_ack` succeeds, the poller's `send_ack` re-reads the mail as raw MIME and needs Gmail's own topmost `Authentication-Results` reporting `dmarc=pass` for the From domain (`fact_confirmation.sender_authenticated`, fail-closed). The printed headers the poller reads never carry that header, so without the second read a forged From would draw the acknowledgement to whoever it names — backscatter from the Executive's mailbox. An unauthenticated sender makes `send_ack` raise `roster_intake.AckWithheld`: the claim is released, and they are held silently. `authenticated_by_gmail` strips quoted strings and comments before reading the verdict — Gmail copies the envelope sender and DKIM tags into its own stamp, and a quoted local part could otherwise carry a fake `dmarc=pass` — and needs exactly one `dmarc=` verdict reading exactly `dmarc=pass header.from=<domain>` (Gmail may follow it with its own `dara=` resinfo, so its position is not checked). Residual: when Gmail writes no verdict at all (a From domain without DMARC) and echoes an unsanitised `;dmarc=pass header.from=…` elsewhere in its stamp, that echo would read as the verdict; bounded by the acknowledgement caps. A withheld acknowledgement is audited `roster_ack_withheld`. `intake` acknowledges before surfacing the card and telling the principal, and both follow the request's `ack_sent_at`: a withheld, failed or capped acknowledgement reads \"haven't been told anything yet\" on the Today card, in the principal's prompt and in the email `[POLICY]` notice.\n\n**Telling the principal.** A new request puts a private card on the principal's `/today` (see Today) and pushes a prompt on their delivery plan — preferred chat DM, then the others, then email — for a chat sender or an email sender on the company's own domain; an outside email sender waits on the card, as does everything past `ROSTER_REQUEST_DAILY_CAP` (30) pushes a day (the same cap bounds how many requests open a day; a declined sender is not raised again for 30 days). A chat prompt can be answered in place (\"that's Annamarie, add her to the team\"). An email prompt carries a one-time token (`RR-` + 20 base32 characters, stored hashed); the poller checks every inbound mail for one **before** triage (`try_email_roster_answer`): the reply must come from the principal's primary address exactly (not an alias, not the company-domain rule), carry the token of a pending request, not be an automatic reply, and not fail DMARC where the header is visible; and is parsed by `parse_decision`'s `roster_identity` shape — anything else falls through to the ordinary path. The token is hidden (`RR-[hidden]`) in every other Google Workspace result the gateway returns — the confirmation sits in the Executive's Sent folder, and any turn could otherwise read it — except the poller's own fetch of an inbound message (`mcp_gateway.reveal_roster_tokens`); a token-bearing mail that is not an answer is scrubbed before triage.\n\n**Replay.** Once the principal answers (web card, chat tool, email reply) — or a People-page / `upsert_person` write makes the sender match someone (`after_roster_write`) — the held messages are replayed through the adapter that received them (`register_replayer`: Slack re-runs the held event, Discord and Telegram their held text, email re-fetches the message from Gmail and handles it as mail from a known sender), oldest first, now attributed to the new Person. A teammate gets every held message; a new contact only held email (it takes the private contact path; held chat is dropped); a decline drops everything. Replays run in a **fresh `contextvars.Context`**: the answer may have come from the principal's own verified turn, and a stranger's message must never be read with the principal's session — and their contacts — still bound.\n\n### Standing facts by email (confirm by reply)\n\nThe principal can ask for a standing fact, a retirement or a company-profile edit by email, and the fact tools check it exactly as in chat (the own-words gates run on `_email_memory_text`, the sender's new text, not the quoted chain or attachments). An email's From line proves nothing, though, and a standing fact is read by every later prompt as the principal's own account. So an emailed change is **held, never applied**, until the principal's own mailbox confirms it (`integrations.fact_confirmation`, `memory.facts.fact_confirmations`):\n\n- **Who may ask.** The poller stamps every email turn with `Session.email_from` (the bare lowercased From address) and `email_authenticated`. The text the poller reads never carries `Authentication-Results` (workspace-mcp prints a fixed set of headers), so for mail claiming the principal's primary address it reads the message again as raw MIME (`fact_confirmation.sender_authenticated`) and requires Gmail's own stamp: the **topmost** `Authentication-Results` header (Gmail adds it on receipt, above anything the sender wrote), authserv-id `mx.google.com`, reporting `dmarc=pass` for the From address's domain, with the raw From matching (`authenticated_by_gmail`). It fails closed: no header, `dmarc=none`, a temporary error or a failed read all leave the turn unauthenticated. Mail from a teammate other than the principal is read the same way, before its inbound audit row is written: authenticated, the whole handling of it is theirs alone in the audit log (`rows_for_person`, see Audit) and the turn reads their own mail as them (`Session.email_sender_authenticated`); unauthenticated, it keeps the ordinary rule. Mail from anyone else costs no second read and is never authenticated. `fact_tools._gate` admits an email turn only from the principal's **primary** address, exactly (no alias, no company-domain rule), authenticated, and not private to the principal (mail they forwarded stays refused, as do contacts, teammates and unknown senders).\n- **The hold.** The tool returns `awaiting_confirmation` (no action chip, and the model is told to say nothing has changed). `request_confirmation` stores the change with the SHA-256 of a fresh one-time token (`FC-` + 20 base32 characters, 100 bits; 7-day expiry; at most 10 waiting at once) and emails the token to the principal's roster address, outside any turn's outbound context. A profile edit is dry-run first, so one that cannot apply is reported in the turn, before any email goes out. If the email fails, nothing is held.\n- **Keeping the token from any model.** The tool result never carries it, and `mcp_gateway.hide_roster_tokens` now scrubs `FC-` tokens as well as `RR-` ones (`FC-[hidden]`) from every Google Workspace result except the poller's own fetch of an inbound message. Act as me's reads of the principal's own Gmail, where the confirmation email lands, hide them too (`delegation.gmail.parse_message` and the thread search), so the drafting model never sees a live token either.\n- **The reply.** `_handle_one_email` hands every message to `try_email_fact_confirmation` right after the roster-answer check, before any model turn. It acts only on a reply from the principal's primary address carrying the token of a **pending** confirmation. A made-up, spent, cancelled or expired reference matches nothing and falls through to the ordinary path with the token hidden, so no old or forged token can make the Executive email the principal. A reply with a live token is read again as raw MIME (`read_raw`) and must pass the same Gmail DMARC check as the request; one that doesn't (or whose raw read fails) is refused, the change keeps waiting, and the principal is told. Then an **automatic reply** is consumed quietly — no answer, no model turn, the change keeps waiting: the printed headers catch a Gmail vacation reply (`Precedence: bulk`), and the raw headers catch what they don't show (`automatic_reply`: `Auto-Submitted` other than `no`, `X-Auto-Response-Suppress` and similar, `Precedence: auto_reply`, an out-of-office subject), since an Exchange out-of-office echoes the token from the principal's own address and may well say \"I'll confirm on my return\". The raw message starts at the tool's exact separator line (`\\n\\n--- RAW MIME ---\\n`), so a sender-written Subject imitating it cannot supply the headers. CONFIRM (or yes / approve) applies the change through `fact_tools.apply_confirmed`, the same apply step chat uses. CANCEL (or no) drops it, and an unclear reply leaves it waiting. The principal gets a one-line result either way. A compare-and-set on the row means two replies can apply it only once.\n- **Audit.** Private `fact_confirmation` rows record held / applied / cancelled / refused. The applied change also writes the fact tools' usual private rows, with `source_channel=\"email\"` as its provenance.\n\n### Checking a channel is set up\n\nSettings → Setup status (`GET /setup/status`, `api/setup_checks.py`) gives each channel a light and a plain-language fix. It tests the credentials with one free read-only call — Slack `auth.test`; Telegram `getMe`, then `getWebhookInfo` (is a webhook registered, is it this app's `/webhook/telegram`, did a delivery fail in the past day); Discord `users/@me`, only while the bot is still connecting — and reads what the lifespan left on `app.state`: the Slack socket's `is_connected()`, the Discord bot's `is_ready()` together with its gateway socket being open (discord.py stays ready while it reconnects after a drop), or the exception its task stopped with (`LoginFailure`, `PrivilegedIntentsRequired`), and the MCP gateway. A listening Slack is amber while the listener hasn't resolved its own user id (`slack_bot.bot_user_id()`), since thread continuations are off until it does. Because inbound is roster-gated, a connected Slack, Discord or Telegram is amber while nobody on the roster has its id (`slack_user_id`, `discord_user_id`, `telegram_chat_id`), or when its latest `integration_inbound` row is a `Rejected: …` from the past day whose sender still isn't on the roster — the page shows the sender's id, never message text. Local login with no `TELEGRAM_WEBHOOK_SECRET` is red for Telegram, since the local-login gate turns its webhook away. Google Chat and Gmail are checked from configuration: a numeric project number and a readable service-account key; the OAuth client (or service-account key) the workspace-mcp child signs in with. Gmail is also amber when startup priming could not find every pinned Google tool (`MCPGateway.prime_pinned_tools`, whose result the gateway keeps on `pinned_tools`): the row names a few of the missing tools and counts the rest, and points at the `MCPGateway` log lines. It is amber too while that probe has not finished, and when the probe itself failed — the one check that reads green is the one that ran and found everything, so a crashed probe is never silently fine. The row speaks for the whole `google_workspace` server, so a missing Docs or Drive tool shows up under Email (Gmail).\n\nA channel setting written `KEY= # note` in a `.env` file arrives as the note's text, because python-dotenv strips only a comment that follows a value. `Settings` reads such a value as unset for the Slack, Telegram, Discord and Google Chat credentials, so the line no longer switches a channel on with its own comment as the token. `TELEGRAM_WEBHOOK_SECRET` is the exception, because read as unset a junk secret would turn the webhook's check off. It is kept instead, and since Telegram can only ever send 1–256 of `A-Z a-z 0-9 _ -` (what `setWebhook` accepts), a secret outside that set — a leftover comment, a pasted `<value from Step 2>` — could only be matched by someone who guessed it: while it is set the webhook refuses every update, neither the local-login gate nor the roster tools count it as proof (`Settings.telegram_webhook_secret_valid`), and Setup status shows it red (`.env.example` shipped four such lines; the notes now sit on their own lines).\n\n### Conversation state\n\nEvery adapter keys a deterministic `session_id` off the conversation surface and replays that session's stored turns into `Session.conversation_history`, so a follow-up (\"yes, go ahead\") lands with the question still in context. Slack's ids are `slack:dm:{user}`, `slack:thread:{channel}:{thread_ts}`, and `slack:channel:{channel}:{user}` for a mention that starts a thread; because the bot's reply roots a Slack thread at that message's `ts`, the turn is double-written to that thread's session too, so the first continuation inside it is not cold. A reply inside a thread THAT PERSON ROOTED also offers the parent `slack:channel:{channel}:{user}` id when matching an approval gate, since a gate raised on the mention recorded that session while the answer arrives under the thread's. The root-author check is load-bearing: widening on \"threaded and not a DM\" alone offered one person's channel session inside any thread in that channel, so a remark they made to a colleague could resolve their open gate and post the run's title into that colleague's thread. Discord has the same shape — its auto-thread router promotes a mention reply into a bot-owned thread — and the same alias. An @-mention inside a DM is handled once, by the DM path: Slack fires both events, and taking both forked the conversation across two session ids. The same id is the audit grouping key and the scope passed to `format_for_prompt`. A per-session `asyncio.Lock` serializes turns within one conversation — the inflight semaphore caps fan-out across the workspace, which does nothing to stop two messages in the same thread from both reading history before either writes it. Persistence happens after the reply is sent and its failure is logged, never surfaced: the user already has their answer.\n\n**Conversations within a chat app.** A Telegram chat, a Slack or Discord DM and a person's @mentions in one channel (`slack:channel:…`, `discord:channel:…`) are each one endless stream, so they are cut into conversations (`memory.conversations`, shared helpers in `integrations/conversations.py`). Each adapter computes the stream's base id as before and asks `conversation_for` which conversation a message belongs to: the newest, unless the stream has been quiet for two hours (`CONVERSATION_GAP`), when the next one starts. The first conversation keeps the bare base id; later ones add `~2`, `~3`, … (`memory.conversation_ids`). A whole message of `/new`, \"new chat\" or \"new conversation\" starts a fresh one instead of a turn and is answered with the name of the one it ended; the new row has no messages and stays off the Chats page until its first turn. After a conversation's first exchange `title_after_turn` names it in the background with `utils.session_title.generate_session_title` (the first words of the message if that fails), as a web chat's first turn does. Threads (`slack:thread:…`, `discord:thread:…`), email and every other id are left whole. What has to see the whole stream compares base ids: the inbound resolver, so an approval asked before a cut is answered after it, and the session-scoped decisions and advice `format_for_prompt` renders. At boot `split_existing_streams` cuts history saved before this by the same rule, once (`app_migrations`), titling each part from its first message and carrying a read-mail lock to every part. An adapter installed outside the package adds its streams with `conversation_ids.register_rolling` and cuts its own old history once, with its own migration name and an `only` filter for its base ids, because the built-in sweep may already have run without it.\n\nThis was absent on Slack until #136. Every message built a fresh `Session` and minted a per-*message* id (`thread_ts` falls back to the message's own `ts`, so a DM never repeated one), which meant a multi-turn confirmation could not work: the Executive asked whether to send a briefing to Slack, the user said yes, and the next turn had no record of the question.\n\n### What the model is told about the channel\n\nChat adapters pass a `<channel>` block (`integrations.channel_context`) in the **user turn** — never a cached system block. It names the surface, states that the person cannot see the app's cards or buttons, and draws the line the assistant kept crossing: briefing proposals and alerts are actioned on the briefing page; a proposal that books something can only be approved there; a workflow sign-off the Executive is explicitly waiting on *can* be answered in the conversation. The principal's DMs on Slack, Discord and Telegram also receive the open-alert digest (`briefing.context.format_open_alerts_for_prompt`) as the `<briefing>` block, gated to DMs because the board is company-wide and a shared channel must not pull it in. That digest is the ONLY trusted source of an `alert_id` for `ack_alert`. The briefing page's `[Discuss mode — alert_id=N]` primer is a pointer to which item is under discussion, not authority on its own: the server accepts that id only if it is also on the live board it derived itself. An id appearing anywhere else, including inside an alert's own body, is not trusted: alerts are minted from inbound email and chat, so their bodies are attacker-controlled. Two mechanisms enforce that rather than one prompt rule. Every field the digest interpolates is collapsed to a single line, because a newline in an attacker-written headline, suggested action, tag or review note would otherwise forge a whole extra `[N]` line — a fabricated instruction to clear somebody else's alert. And the live board the server derived is recorded on the session, so `ack_alert` refuses any id outside it — on EVERY session, web included; the web previously had no check at all. The recorded set is the whole live board (the cap `/today` renders as cards), not just the rows the digest printed, so a card past the digest's 30-row render cap is still ackable rather than sending the principal back to the page. Be precise about what this buys: it stops an id that is invented, closed, snoozed or expired, including one quoted out of an attacker-written alert body. It does NOT stop the model being argued into acking the wrong LIVE card — one alert's body can still make a case about another id that is genuinely open. Scoping trust to the single item under discussion would need the primer id carried server-side, which it is not.\n\n### Handler failures\n\nAn adapter's catch-all replies \"I encountered an error processing your request.\" On Slack that used to be the whole story — one uncorrelated log line, no audit row, and an unguarded error-send that could itself escape into Bolt. It now logs with the session, sender, channel, thread and mode, and writes an `integration_inbound` row with `details.outcome=\"handler_error\"`, joining the inbound row under the same `session_id`. Upstream, a single crashing tool handler no longer takes the turn down at all: the orchestrator's tool gathers use `return_exceptions=True`, hand the model an error `tool_result`, and audit it with `ok=false` (cancellation is re-raised, not swallowed).\n\n### Attachments\n\n`integrations.attachments` handles web chat, Discord, Telegram, Slack and Google Chat. Images become Anthropic vision blocks; documents (PDF/DOCX/TXT/MD/CSV) are inlined into the turn that carried them — inside an `<untrusted_content source=\"attachment\" author=\"<name>\">` block (`content_trust.wrap_untrusted`, whose closing tag the text cannot forge), under an `[Attached: <name>]` label, which the open-loop pass reads as \"not the speaker's words\" and the decision extractor strips before it reads the turn — and indexed into the ChromaDB `inbound_attachments` collection. 20 MB/file cap; web chat uses `POST /chat/upload`.\n\n**Scanned PDFs** are read, not reported empty. A PDF with no text layer goes through `knowledge.pdf_reader`: with `PDF_PROVIDER_READING=true` (opt-in, off by default) the deployment's own model reads it through its provider's PDF support (Anthropic natively, OpenRouter's `file-parser`, a local server only with `LOCAL_PDF_INPUT`); otherwise, and as the fallback, local OCR (pypdfium2 + RapidOCR, offline, no key) — so every deployment converts them. The label then reads `(converted from scanned pages)`, and an unreadable one says why. The index takes the text already extracted for the turn, so a scan is converted once. A file a channel delivered (everything here except `POST /chat/upload`, where the signed-in user sends it) converts at most `PDF_INBOUND_MAX_PAGES` (30) pages and draws on a rolling `PDF_INBOUND_PAGES_PER_HOUR` (300) budget shared by all senders, so sending scans cannot run up unbounded model spend or CPU.\n\nPer channel:\n- **Slack** — a `file_share` message (previously dropped outright for its `subtype`) is handled like any message; files download only after the roster gate and the response gate, with the bot token sent only to `https://files.slack.com/` (needs the `files:read` scope). History and peer memory record the words plus `(Attached files: …)`, never a document's text.\n- **Google Chat** — uploaded attachments download through `media.download` with the reply credentials (`chat.bot`), checked against the 20 MB cap per chunk, 5 per message; an attachment-only message is no longer dropped. The webhook authenticates Google, not the sender, so files are read only when the event's `user.email` is on the People roster (a team member); anyone else's files are named to the Executive, never downloaded. A Drive file shared into a message is left to the Drive tools, with its id.\n- **Email** — the poller (`integrations.email_attachments`) downloads up to 5 readable attachments of the message itself via `get_gmail_attachment_content` — workspace-mcp in stdio mode saves the file and answers `Saved to: <path>`, which must start its line, be the last such line and end in the answer's `Saved filename:` (the answer echoes the sender-chosen attachment name first, and a name carrying `Saved to: /…` must not redirect the read), then resolve inside the tool download folders — and appends their text to the turn (15k chars each). Only for a sender the principal knows (team, contact, principal); a stranger's files stay a list. It runs outside the model loop, so it works on private turns without adding the attachment tool to `PRIVATE_TURN_MCP_TOOLS`.\n- Anything else (a Drive PDF, a skipped attachment) — the Executive's `read_document` tool.\n\nIndexed chunks are keyed on `attachment:<bare filename>` and tagged `type=attachment`. An attachment's name is chosen by whoever sent the message and the name is the chunk-id namespace, so without the prefix an inbound `strategy-2026.md` would upsert over the curated company document of that name. The prefix separates inbound content from curated uploads and keeps provenance visible; it does not separate senders from one another.\n\nIsolation is the **collection**: `retriever.retrieve` never queries `inbound_attachments` at all, so an attachment cannot resurface as company knowledge in a later, unrelated turn. Attachment processing is roster-gated on Discord too — it previously ran before the gate, so an unrostered sender's file was downloaded and indexed before their message was dropped. These rows still have no operator surface: they are never written to `company/docs/`, so `GET /documents` does not list them.\n\nThis replaced a domain-based scheme that did not work. The chunks lived in `company_docs` under a domain outside the specialist set, on the premise that an unrecognised domain matched no filter. It matched every one — `ChromaDBStore.query` attaches a `where` clause only when a domain filter is present, and an Executive-level retrieval passes none — so attachments came back under “From your company documents”, top-ranked and indistinguishable from a curated upload. Legacy rows are migrated across on boot.\n\n### MCP gateway & outbound\n\nMCP runs in **both directions**. As a *client*, `orchestrator.mcp_gateway.MCPGateway` proxies MCP servers (Gmail, Calendar, Drive, Outlook mail and calendar — whatever `mcp_servers.json` configures) as Executive tools, with egress gating on Gmail recipients, Calendar attendees, Drive share grantees, and Microsoft 365 recipients/attendees against the roster (and a refusal of public / whole-domain Drive sharing). The Gmail gate also accepts only a fixed set of argument keys, so workspace-mcp's `reply_all` (recipients taken from the thread, never checked) and `forward_message_id` / `include_forwarded_attachments` (forwarding any inbox message) are refused; a reply is `thread_id` plus an explicit, roster-checked `to`. Three more refusals close paths around those gates: every Apps Script tool (code running as the Executive's account, outside the gates, and since workspace-mcp 1.29.0 schedulable to keep running); any argument that makes the server fetch a URL (`fileUrl` / `file_url` on the Drive and import tools, a Gmail attachment given as a `url`), since the fetch can carry data out in its query string before any recipient gate matters; and text Google would mail to someone the roster never checked (`rsvp_comment` to an event's organizer, an out-of-office `decline_message` to whoever invites the Executive). Every gate reads its arguments by normalized spelling: workspace-mcp 1.29.0 renames a camelCase key such as `Attendees`, `userGoogleEmail` or `rsvpComment` to the real parameter, so a gate matching the exact key would never see it. Arguments that are not a JSON object are refused for the same reason. Google-side fetches are refused too: any key ending in `url` / `uri` at any depth whose value has a URL scheme (a Slides image or page background, a Docs or Forms image; a document hyperlink and a meeting link are stored, not fetched, and pass, as does a Drive file id), (a value is a URL after the characters a URL parser strips are removed, and `//host` counts), and on every tool that can write cells — the Sheets tools and the Drive create / update / import tools, which turn CSV into a Sheet — any string anywhere holding a formula call that fetches (`IMPORTDATA`, `IMPORTXML`, `IMPORTHTML`, `IMPORTFEED`, `IMAGE`), JSON-encoded or mid-CSV included: a string argument that encodes JSON is read as the server decodes it (one too large to read is refused), text is also scanned with character references decoded and comments / CDATA removed, and an upload that would become a native Sheet may only be text `content` — a `base64_content` or server-side `file_path` upload there is refused, since an XLSX hides its formulas behind zip members and XML encodings; a PNG to Drive or a DOCX to Docs is untouched. Not covered, unverified: whether Drive's HTML / Markdown import fetches an `<img src>` when it converts `content` into a Doc. A `manage_event` call that names no attendees is sent with `send_updates=\"none\"`, so the event's existing guests are not emailed text the gate never checked. The allow-list (`_egress_sets`, built by `_roster_allow_set` as a `people.identity.RosterAllow`) is every team member's email and email aliases plus the Executive's own address, always, and — for a teammate — any address on one of the company's own domains whose local part (`+tag` dropped) matches theirs, the same rule inbound mail is matched by; a **contact's** email joins it only when `people_tools.contacts_reachable_now()` is true — a turn the principal started on a verified surface (web chat, their own Slack or Discord, a verified private Telegram chat), or inside `grant_contact_egress()` (the decisions approve route, when the approver is the principal). On any other turn — an inbound email, a teammate, the scheduler, a workflow, the alert review — a contact's address is refused exactly like a stranger's, so the refusal cannot reveal that it belongs to a contact (contacts are private to the principal). The single exception to the roster rule is the fixed roster-request acknowledgement above, under its one-shot `roster_ack_grant`. A turn about the principal's private mail (`Session.private_to_principal`) narrows the list to the principal's own address and the Executive's. Such a turn may call only the MCP tools in `schedule_tools.PRIVATE_TURN_MCP_TOOLS`: Google Workspace reads (`search_gmail_messages`, `get_gmail_message_content`, `get_events`, `list_calendars`, `query_freebusy`, and the Drive, Docs and Sheets reads such as `search_drive_files`, `get_drive_file_content`, `get_doc_content` and `read_sheet_values`) and the two Gmail tools this gate narrows to the principal (`send_gmail_message`, `draft_gmail_message`). It is an exact allow-list, not a server: workspace-mcp runs at the tier `WORKSPACE_MCP_TOOL_TIER` names, `complete` by default, where a Google Chat message to a space, a Docs, Sheets or Drive write into a file already shared with others, `manage_event` on a shared calendar (its gate checks attendees and pins notifications, nothing else) reaches people with no recipient check; and every other server has none at all (a Slack or Notion server, the default `fetch` server, whose URL can carry the turn's content, a server loaded at runtime). Its `search_tools` results show only the allowed tools (`workflows.tool_catalog.filter_search_results`; output it cannot parse reads as no match), and a `call_tool` naming anything else is refused with an error tool_result and an audit row private to the principal (`schedule_tools.private_turn_withholds`), a missing name included. `search_tools` and `call_tool` stay offered, so the tool list and its cached prefix do not change. The typed calendar tool (`_resolve_attendees`) and the chat DM tools (`message_person`, `send_telegram_message`, `send_discord_dm` and their person-id recovery) apply the same rule — a contact's id reads as an unknown one off the principal's turn — and a private turn may DM or invite only the principal (`PRIVATE_TURN_REFUSAL` otherwise); `message_person` never falls back to a \"Needs you\" alert for a contact, who cannot sign in to see it. The Google Workspace server (`workspace-mcp`) behind those Gmail/Calendar/Drive tools runs **co-located inside the API** as a stdio child of the gateway — fitting the single-tenant model (one install = one company = one exec Google account), so there's no separate service to operate. It's baked into the API image and launched by `docker/workspace-mcp-launch.sh`, which selects single-user OAuth (token persisted at `/data/google_credentials`) or service-account domain-wide delegation from `GWORKSPACE_AUTH_MODE`. The optional `WORKSPACE_MCP_TOOLS` (comma-separated service names, default `all`) limits which Google services it loads. The gateway forwards the Google secrets to the child via `_FORWARDED_ENV_VARS`. Every `google_workspace__*` call acts as the Executive's own account. workspace-mcp takes the acting account per call (`user_google_email`) and nothing used to check it, so under domain-wide delegation a model-chosen address could open anyone's mailbox in the domain; `mcp_gateway._check_acting_account` now refuses a value naming any other address — before every other gate, with an `integration_outbound_blocked` audit row. An absent value stays allowed (the server's own account).\n\n**What the model is told is connected.** `MCPGateway.start()` records the configured server names (`server_names`), and the Executive renders them with the channel settings into block 0's *Connected Systems* section (`prompts.connected_systems`; see the caching section). While Google is connected that section pins the Google tool names the Executive uses constantly (`GOOGLE_TOOL_MANIFEST`, narrowed by `connected_systems.google_tool_manifest` to the services and tier actually loaded) and tells it to call them with `call_tool` directly, with no `search_tools` round trip. The narrowing is the point: a name pinned for a service workspace-mcp never imported is not in the index, so the model calls it, receives extensible-mcp's “has not been discovered via search_tools” refusal, and has no way to square that with its own instructions. extensible-mcp only runs a tool one of the session's searches has returned, and remembers it for the session's life, so the gateway does that discovery itself: the lifespan starts `MCPGateway.prime_pinned_tools()` in the background once the gateway is up with `google_workspace` configured (an exact-name search per pinned tool, the way `workflows.tool_catalog.resolve` does it, over the same narrowed set the prompt advertises — `connected_systems.pinned_google_tools`, so priming never probes a service that was not loaded). A name neither the prompt nor priming can account for is logged once, naming both causes: excluded by `WORKSPACE_MCP_TOOLS` / `WORKSPACE_MCP_TOOL_TIER`, or renamed by workspace-mcp since the pin. A pinned call that still comes back with extensible-mcp's \"has not been discovered via search_tools\" refusal, for example before priming finished, is discovered and retried once. That happens after every gate has passed the same arguments, and only for pinned names: any other tool still needs the model's own search, and nothing here widens what a call may do, including the private-turn allow-list.\n\n**Microsoft 365** is the Outlook twin: `@softeria/ms-365-mcp-server` (a Node package, pinned) is installed in a throwaway Node build stage and only the `node` binary plus the package are copied into the API image, launched by `docker/ms365-mcp-launch.sh` as the `microsoft_365` entry with an **explicit tool allow-list** (`--enabled-tools`, an anchored regex) rather than the upstream `mail,calendar` preset — that preset also registers inbox forwarding rules, mailbox settings (external auto-reply), calendar sharing and event forwarding, which can address people the gateway never sees. Exactly the Outlook mail + calendar tools and the OneDrive tools are exposed (`microsoft_365__send-mail`, `microsoft_365__create-calendar-event`, `microsoft_365__search-onedrive-files`, …; hyphenated Graph endpoint aliases, proxied verbatim), and every admitted tool is either roster-gated or carries no recipient; the gate also covers the excluded writers in case an operator widens the list. Both servers can be configured at once. Auth is the MSAL device-code flow against the operator's own Entra app registration (`MS365_MCP_CLIENT_ID` is required; the server's built-in public client is refused); the refresh token persists at `/data/ms365_credentials` and is seeded once with `ms365-mcp-launch.sh --login` from inside the container — the server has a native device-code CLI, so there is no mint script and nothing to copy. The launcher scrubs the literal `$VAR` placeholders extensible-mcp leaves for unset variables (otherwise MSAL would be handed `$MS365_MCP_TENANT_ID` as an authority), and `MS365_MCP_EXPECTED_USERNAME` pins the bound mailbox. The egress gate for this family walks Graph's nested shapes (`body.Message.toRecipients[].emailAddress.address`, `body.attendees[]…`) — every email-shaped string outside the free-text fields must be on the roster, control characters and non-ASCII addresses are refused, unforeseen keys are still scanned (fail-closed) — and covers the whole write surface: send, reply, reply-all, forward, every draft tool, `update-mail-message` (it can PATCH recipients onto a draft) and create-/update-calendar-event; reads and `delete-calendar-event` (the only cancel path) pass through. Graph's reply, reply-all and send-draft actions carry no recipient in their arguments (the recipient is whoever the referenced `messageId` names), so for those the gate first reads the referenced message through the server and roster-checks the recipients Graph will derive (`replyTo` else `from` for a reply; `replyTo` and `from` together, plus the original to/cc, for reply-all; the draft's own to/cc/bcc for send-draft), refusing on any lookup failure; RSVP and cancel actions get the same treatment through an event lookup, because their comment is emailed to the organizer or every attendee; `download-bytes` is pinned to mail attachment bytes and OneDrive file content (`/drives/{id}/items/{id}/content`, optionally Graph's `?format=pdf` rendering), and a OneDrive result comes back as the file's extracted text rather than base64 — the attachment pin is what makes the poller's promise that an unrostered sender cannot be auto-replied to hold on Outlook as it does on Gmail; and the rendered inbound text quotes any forged `--- REPLY ---` line inside a body so the persona only ever reads the real block. Every mail write's `from`, `sender` and `replyTo` may name only the Executive's own address (the Outlook twin of the Gmail acting-account pin), so the model cannot send as a rostered colleague's mailbox or steer replies elsewhere; and `move-mail-message` into Deleted Items, Junk Email or Recoverable Items is refused, so a move cannot trash mail ungated. The server registers its account tools whatever its tool allow-list says, so the gateway itself refuses `login`, `logout`, `select-account` and `remove-account` (audited like any other refusal) and drops them from `search_tools` results — a chat turn cannot sign the mailbox out, switch accounts or start a sign-in; setup is only the launcher's `--login`, and the read-only `verify-login` / `list-accounts` stay callable. The example config's hyphenated patterns still deny `delete-mail-*`, `logout`, `remove-account` and `download-bytes-to-file` as defense in depth. A successful `send-mail` records the same outbound-context linkage Gmail does.\n\n**OneDrive in chat** (opt-in with `MS365_MCP_ONEDRIVE=true` and a fresh `--login`, since the server asks for its tools' permissions on every token refresh and a sign-in without file consent would otherwise fail mail and calendar too) matches what the Executive can do on Google Drive: search, list a folder, read a file, upload and edit, create folders, move, rename, copy and share; there is no delete tool. The launcher adds `Files.Read.All` at sign-in (`--extra-scopes`), so folders shared *with* the Executive can be read too; writes (upload, new folder, move, copy) land only in the Executive's own drives, checked against `list-drives` at call time, so a planted instruction cannot save company files into a stranger's shared folder. Sharing is gated like Drive sharing: `share-drive-item` invitees must be People on the roster and named by email (an `objectId` or `alias` grantee is refused, since nothing can check it), no `owner` role, and `create-drive-item-share-link` must be scoped to specific people (`scope: \"users\"`): an anonymous or whole-organization link, or one with no scope (the tenant default), is refused. An invite must require sign-in and grant only view or edit, a link must be a view or edit link, and `search-onedrive-files` takes plain words only, since its query goes into the request path unencoded.\n\nAs a *server*, `openexecutive.mcp_server` exposes OE's own company context and specialist council to external MCP clients at `/mcp` — the inverse direction; see the dedicated **MCP Server** section. The gateway launches extensible-mcp as a `uvx` subprocess whose fastembed tool-search index uses an embedding model baked into the API image and loaded offline, so it no longer re-downloads that model from the Hugging Face Hub on every cold start. extensible-mcp is pinned twice over: to a commit, which fixes its own code, and with `--exclude-newer`, which fixes the resolution of its dependencies — without the cutoff uvx re-resolves them against PyPI on every networked start, so a new release of any of them would reach running deployments on restart. Both live in `_EXTENSIBLE_MCP_LAUNCH_ARGS` in `orchestrator/mcp_gateway.py`, the Dockerfile's pre-warm uses the same arguments (a unit test enforces the match), so an image whose pre-warm succeeded starts the gateway with no network. The pre-warm is best-effort (`|| true`): an image built while GitHub was unreachable fetches the pinned commit on its first start instead. Updating the gateway is a deliberate bump of the commit and cutoff; `uv` itself is not pinned. Because the MCP stdio client does not inherit the parent environment, `MCPGateway.start()` explicitly forwards the embedding-cache/offline env vars (`FASTEMBED_CACHE_PATH`, `HF_HUB_OFFLINE`, …) into the child.\n\n**Emailing an artifact.** A Gmail send or draft may carry `attachments=[{\"artifact_id\": \"alert:12\", \"as\"?: \"docx\"}]`. `call_tool` runs the recipient gate first; only if it passes does `_expand_artifact_attachments` replace each such entry with the rendered file (`content` base64 + `filename` + `mime_type`, workspace-mcp's own attachment shape), produced by `artifact_records.render_artifact_file` — the same renderer as `/artifacts/{id}/download`. Only artifact rows resolve (never an arbitrary alert), links have no file, and an entry mixing `artifact_id` with other keys is refused. A stringified or single-object `attachments` is normalised first, so an artifact entry can never reach workspace-mcp unexpanded (it would silently drop it). At most 5 distinct artifacts per email (repeats are dropped) and 15 MB rendered in total; rendering runs off the event loop via `asyncio.to_thread`. A refusal writes `artifact_attachment_refused`; a successful send writes `artifact_attached` — both carry the recipients. Other attachment entries pass through unchanged. The same expansion serves the Microsoft 365 mail writes: the entry becomes a Graph `#microsoft.graph.fileAttachment` (`name`, `contentType`, `contentBytes`) placed where Graph expects it for that tool (`body.Message.attachments` for `send-mail`, reply, forward and the createReply drafts; `body.attachments` for `create-draft-email`), whether the model put it at the top level or inside the message; an entry anywhere else, one on `update-mail-message` or `send-draft-message` (their requests cannot carry attachments), or one behind a body that is not an object is refused rather than forwarded unexpanded.\n\n**Enablement, and why it degrades instead of dying.** `MCP_ENABLED` left unset is inferred from the presence of `mcp_servers_config_path` — dropping an `mcp_servers.json` next to `profile.yaml` turns MCP on with no further configuration. An **explicit** setting always wins in both directions: `MCP_ENABLED=false` keeps MCP off with the file still in place. It did not: `_resolve_mcp` tested `not self.mcp_enabled`, which cannot tell \"never set\" from \"set to false\", so the file's presence silently overrode the operator. The file is probed through `config.mcp_config_file_present`, never `Path.exists` directly — that re-raises an EACCES on a parent directory, and from inside a validator that comes out of `Settings()` itself, which `api/main.py` calls at module level. The check now reads `model_fields_set` (the fields the env actually supplied) and stores that on a private attr behind the read-only `mcp_auto_enabled` property, because the validator's own assignment joins that set and would otherwise make every auto-enable look explicit downstream.\n\nStartup is **best-effort**, in `main._start_mcp_gateway`. Two checks come before the subprocess: MCP off is a no-op, and a config defining no servers under `mcpServers` is logged and skipped — extensible-mcp's own loader raises on that and exits, and because the child is already gone the only thing that surfaces is anyio's \"Attempted to exit cancel scope in a different task\" from `stdio_client` unwinding, a traceback naming nothing about configuration. The start is also bounded (120s, generous because that child may still clone extensible-mcp on a cold start if the image's pre-warm failed): `ClientSession` carries no `read_timeout_seconds`, so `initialize()` waits forever on a child that is *alive but silent*, and an unbounded wait is a lifespan that never yields — the same healthcheck-never-passes symptom, which no `except` can see. A gateway that fails anyway is logged at ERROR with the config path, the servers it named, and whether the config file or an explicit `MCP_ENABLED` turned MCP on, then torn down under its own timeout; the API boots **without MCP tools and without the email poller** rather than refusing to boot. The guard catches `BaseException`, not `Exception`, re-raising a bare `CancelledError` first: the original failure came from anyio, which wraps a task group's failures in `BaseExceptionGroup` as soon as one is a `BaseException` — and that group is not an `Exception`, so `except Exception` would miss the very shape this contains. The teardown is contained too, since it runs from inside that `except` and would otherwise replace the actionable log with the crash it prevents. `configured_server_names` never raises either, and deliberately catches broadly: it reads an operator-owned path during startup, where an exception is the crash loop rather than a bug report — `Path.is_file` re-raises an EACCES on a parent directory, and `json.loads` answers deeply-nested input with `RecursionError`, which is not a `ValueError`. It tests `is_file()` so a directory or a symlink to a FIFO is never opened, and caps the read at 1 MiB. That exception used to propagate out of the lifespan and kill the container — under `restart: unless-stopped`, a crash loop whose only outward symptom was a healthcheck that never passed. Outbound tools: per-Person DMs (`send_slack_dm`, `send_discord_dm`, `send_telegram_message`) plus room-targeted `send_department_message` and `send_company_broadcast`. All three per-Person DMs refuse a recipient id that matches no non-archived Person row (`schedule_tools._dm_recipient_on_roster`), which is what stops untrusted content in a turn — inbound mail, a chat message, a knowledge document — from directing a DM at an arbitrary workspace member. `send_slack_dm` previously consulted the roster only when the turn was already private to the principal, so on an ordinary turn it applied no roster check at all while its Discord and Telegram siblings did; the room-targeted tools are deliberately ungated, since those channels are team-visible. The per-Person DMs are additionally rate-limited and de-duplicated by the outbound anti-spam guard (below).\n\n### Outbound anti-spam guard\n\nEvery per-Person DM passes a deterministic guard (`orchestrator.outbound_guard.check_outbound_allowed`) after the roster gate and recipient-id recovery but before the network send — the single chokepoint **all** proactive DM paths reach (ad-hoc chat sends, the scheduler's followup/nudge dispatch, principal-brief delivery, the reflection turn, and research-finding routing). On a violation it returns a human-readable reason and the handler **suppresses** the send (`{\"status\":\"suppressed\",\"reason\":…}`), writes a `tool_invocation` audit row with `suppressed=true`, and writes **no** `done` activity row — so a blocked attempt never counts toward its own rate cap, and the Executive can read the reason and hold or `schedule_followup` for the recipient's next window. Three controls: **content dedup** (a near-identical message — whitespace/case-normalized, first 160 chars — already sent to the same recipient within `OUTBOUND_DEDUP_WINDOW_MINUTES`, default 360); a **per-recipient rate cap** (`OUTBOUND_MAX_PER_RECIPIENT_PER_WINDOW`, default 5, over `OUTBOUND_RATE_WINDOW_MINUTES`, default 60); and **quiet hours / availability** (`OUTBOUND_RESPECT_QUIET_HOURS`, default on — refuse while the recipient is on leave or outside their `people.channel` availability windows, surfacing the next open window). Delivery history is read from the existing `status='done'` `scheduled_actions` rows (`episodic.recent_sends_for_channel_ref`), so there is no new table. The guard is **fail-open**: any internal lookup error returns \"allow\", so it can never turn a working send into a crash or a silently-dropped message. Principal briefs and the internal heartbeat kinds use the `__internal__` channel and never reach these handlers, so they are unaffected.\n\n### Calendar & video links\n\nThe Executive books meetings through two typed tools over the configured calendar backend (`CALENDAR_PROVIDER`: Google Calendar via `manage_event`, or Outlook via `create-calendar-event` — see *Workspace providers*). `create_calendar_event` schedules a **future** meeting (propose→approve by default, or auto-execute once that decision class is promoted) and may be used proactively when a stalled goal warrants a sync. `create_instant_meeting` spins up an **impromptu call starting now**, auto-executing without the approval step and skipping the business-hours/horizon caps, so the Executive can hand back a join link immediately. Because nothing approves it, an instant meeting, and `cancel_calendar_event` too, runs only when someone on the team asked on a surface that verified who they are (the principal, or a rostered teammate in the app or their own chat); a turn an inbound email started gets a refusal to pass on instead, and a team-mode `create_calendar_event` the operations department would auto-execute is proposed instead. Both attach a video-meeting link by default (`add_video_link`; opt out for in-person) — a **Google Meet** link parsed from the created event's `conferenceData`, or a **Microsoft Teams** link from `onlineMeeting.joinUrl` — surfaced to the caller plus persisted on the decision ledger as `meet_link`. The tool descriptions are static, backend-neutral text, so the cached tool block never depends on settings; the pre-rename `add_google_meet` key is still honoured for old ledger rows. Either way the caps that matter still hold — roster membership, principal-protection, max attendees, and a daily booking ceiling.\n\n**Around-the-meeting participation.** The Executive doesn't join the call as a live voice; instead, on every successful booking it schedules a best-effort post-meeting follow-up (~5 min after the end) that DMs a human attendee asking for a quick recap — key decisions and action items. The reply flows back through the same cross-session DM context linkage described below, so the Executive can act on it (summarize, follow up) in a later turn. No meeting-transcript or Drive integration is required.\n\n### Workspace providers\n\nThe Executive can *discover* any mail or calendar tool the gateway exposes, but three code paths call a fixed backend with no model in the loop and used to spell Gmail tool names inline: the inbound mailbox poller (list unread → fetch → mark read), alert email dispatch plus the scheduler's email send-tool hint, and the typed booking tools plus the approval-time conflict check. `integrations.workspace` puts those behind two Protocols (`MailProvider`, `CalendarProvider`) with one module per backend (`google.py`, `microsoft.py`) and a registry keyed on `EMAIL_PROVIDER` / `CALENDAR_PROVIDER` — separate switches (Outlook mail with a Google calendar is legitimate), both defaulting to `google`, so an existing install changes nothing until an operator opts in. The Google backend is the old inline code moved verbatim; tests pin its argument dicts byte-for-byte. The Microsoft backend polls only the Inbox folder (`list-mail-folder-messages` with `isRead eq false`, never the all-folders listing that includes Junk), converts Graph's HTML bodies to text with a real parser, lists a message's attachments so the names appear in the same `N. name (mime, size KB)` shape Gmail's text uses (and reach peer memory), marks read with `update-mail-message`, sends through `send-mail`'s nested shape, books with `isOnlineMeeting` + `onlineMeetingProvider=teamsForBusiness`, cancels with `delete-calendar-event`, and answers the advisory conflict check from the Executive's **own** calendar view (delegated permissions cannot read attendees' free/busy — narrower than Google's `query_freebusy`).\n\nEvery backend renders an inbound message into the same `Subject:` / `From:` / `To:` / `Cc:` + `--- BODY ---` text the persona already expects, and the poller adds a `--- REPLY ---` block naming the exact reply tool and threading ids for *that* message (Gmail: `send_gmail_message` + `thread_id`; Outlook: `reply-mail-message` + `messageId`) after the mail's `<untrusted_content>` block, never inside it. The block rides in the user turn — which is what lets the cached persona say \"use the reply tool in the REPLY block\" instead of naming a vendor tool — and the scheduler's proactive-email trigger names the backend's send tool the same way. Providers never raise on a malformed backend response (they return empty / `None` / `{\"error\": …}`), so one bad cycle never kills a loop, and a chosen backend whose MCP server is missing from `mcp_servers.json` is logged at startup and skipped by the poller: fail-soft, never a boot failure. The poller's newer checks run on that rendered text for both backends (contacts and private turns, roster requests and their emailed answers, the audit row), with three Outlook differences that fail closed: no sender is ever authenticated (the `dmarc=pass` read is Gmail's), so there is no acknowledgement to an unknown sender, no emailed fact confirmation and no principal-authenticated mail; attachments stay a list (the download is Gmail's); and the brief's calendar block is skipped (its listing is Google Calendar's). The code's own emails to the principal (roster prompts, fact confirmations, scheduled deliveries) go through `workspace.registry.send_from_executive`, and `runner.email_ready` follows `EMAIL_PROVIDER`. A private turn's allow-list (`schedule_tools.PRIVATE_TURN_MCP_TOOLS`) and Act as me's mail-touched reads include the Outlook twins of the Gmail and Calendar tools, and roster answer tokens are hidden from Microsoft 365 results as from Google's.\n\n### Cross-session DM context\n\nWhen the principal says \"DM Alex about X\" in one session, OE sends the DM and — if a live conversation is active and the recipient is not one of the principal's contacts (a contact cannot DM the bot, and quoting the principal's conversation into their email reply's unattended turn would put it in the audit log) — records an `outbound_context` linkage row (recipient, full outbound text, originating session id). Email is covered too but recorded one layer deeper — there is no `send_email` tool, so the linkage is written in `orchestrator.mcp_gateway` after a successful `google_workspace__send_gmail_message` or `microsoft_365__send-mail`, one row per `to`/`cc` recipient keyed by the bare lowercased address (OE's own address skipped). When Alex later replies on the 1:1 DM path — or by email — the shared helper `integrations.inbound_hydration.hydrate_user_message` (called by the Discord, Telegram, and Slack handlers, and by the email poller keyed on the sender address) looks up the most-recent `open` linkage for that sender within 72h and prepends an `<outbound_reply_context>` block — the outbound text plus a backstory excerpt from the originating session — to the message handed to `Executive.chat()`. So OE knows what the reply is about instead of starting from an empty per-person session. The block is context for the LLM turn only: peer memory strips it before recording what the person said (`strip_outbound_reply_context`, applied in `sync_turn`, `sync_department_turn` and to the prefetch query), so the Executive's own DM is never attributed to the person who replied to it. The block lands in the user turn (caching rule) and is injected into a copy only — the pristine reply is persisted to history. The linkage is one-shot (consumed on first matched reply, race-safe) so an unrelated later DM isn't re-hydrated. This is deterministic and independent of the per-person `peer_memory` (Honcho) layer.\n\nEvery adapter that persists a turn (Slack, Discord, Telegram, and the web route) records the message's **resolved sender** on `chat_messages.sender_person_id` — the rostered Person the adapter matched, or NULL for an unrostered sender — independently of the session owner, which falls back to the principal on Discord and Telegram so those threads stay visible. Attunement's open loops and feedback rely on this so an outsider's words are never read as someone on the roster's.\n\n### OneDrive folder sync\n\nThe Microsoft twin of the Drive folder sync below (`knowledge.onedrive_sync`, `ONEDRIVE_SYNC_ENABLED`; setup in `docs/onedrive_sync_setup.md`). It copies the files in `ONEDRIVE_SYNC_FOLDERS` (`<drive id>/<item id>` pairs; `openexecutive onedrive-folder <link>` resolves a sharing link) into the isolated `onedrive_docs` collection on its own heartbeat, and shows them on the knowledge page as \"OneDrive\" rows with a **Sync now** button. Microsoft has no service account that sees only what is shared with it (app-only Graph access covers the whole tenant), so it reads as the Executive's own Microsoft 365 sign-in, and the folder list is the scope. It never stores a credential: each tick asks the launcher for a token (`ms365-mcp-launch.sh --access-token`, which runs `docker/ms365-access-token.mjs` through the server's own token cache and will only mint file-read tokens), keeps it in memory, and sends it only to Graph. `knowledge.onedrive_client` issues GETs only, follows a paging link only back to Graph, and fetches a file's pre-authenticated download link without the token; errors never carry that link. PowerPoint and older Office formats are read through Graph's PDF rendering.\n\n### Google Drive folder sync\n\nThe synced files show on the knowledge page's Company documents list as \"Google Drive\" rows (sync time, last edit in Drive, an Open in Drive link and a text view), and the page's source card shows the last run, any sign-in error or count of files that failed to sync, and a **Sync now** button that runs one tick on demand.\n\n\nSeparate from the Drive tools the gateway proxies. The optional folder sync (`knowledge.drive_sync`, `DRIVE_SYNC_ENABLED`; setup in `docs/drive_sync_setup.md`) copies the files in `DRIVE_SYNC_FOLDER_IDS` into the isolated `drive_docs` knowledge collection on a scheduler heartbeat, so the Executive knows those folders without searching Drive live (see RAG). It calls the Drive v3 API directly (`knowledge.drive_client`, `httpx`) as a **service account** with the `drive.readonly` scope (`DRIVE_SYNC_SERVICE_ACCOUNT_FILE`), not as the Executive's own account through workspace-mcp. Google enforces the folder scope, because the account sees only what is shared with it, and the sync can change nothing. Structured JSON means no display text to parse, and every file or folder id is checked against `[A-Za-z0-9_-]{1,128}` before it goes into a URL or a query. The two paths meet in two places. A synced chunk carries its Drive file id and sync time, so the Executive can open the live file with its own `get_drive_file_content`. The setup doc says to share each synced folder with the Executive's account too, so that live open works.\n\n### Confluence space sync\n\nThe optional Confluence sync (`knowledge.confluence_sync`, `CONFLUENCE_SYNC_ENABLED`; setup in `docs/confluence_sync_setup.md`) copies the pages of the spaces in `CONFLUENCE_SYNC_SPACE_KEYS` into the isolated `confluence_wiki` knowledge collection on a scheduler heartbeat, so the Executive knows the company wiki without a tool call (see RAG). It works with Confluence Cloud and Server/Data Center: `knowledge.confluence_client` calls the v1 REST API both serve, directly with `httpx`, signing in with a personal access token (Server/DC) or a username and API token (Cloud). It reads only, never follows a redirect (so the token stays on the configured site), streams responses with a 20 MB cap, and checks page ids and space keys before they reach a URL or a CQL query. The URL must be `https://` unless `CONFLUENCE_SYNC_ALLOW_HTTP` is set, and `CONFLUENCE_SSL_VERIFY` takes `true`, `false` or a CA bundle path for a private certificate authority. With `CONFLUENCE_SYNC_PUBLIC_HOSTS_ONLY` (off by default, for an install where whoever sets the URL must not reach the backend's own network) every request resolves the site's host, refuses it when any address is not public (loopback, private, link-local, shared), and connects to the checked address with the site's name kept for the Host header and the TLS certificate check, directly rather than through a proxy. `confluence_client.config_problem` is the same settings check the app runs at start, for a caller that writes these settings on someone's behalf.\n\nIt is separate from a Confluence MCP server such as mcp-atlassian, which a deployment can add to `mcp_servers.json` for live reads and edits in chat. The two share the env names `CONFLUENCE_URL`, `CONFLUENCE_PERSONAL_TOKEN`, `CONFLUENCE_USERNAME`, `CONFLUENCE_API_TOKEN` and `CONFLUENCE_SSL_VERIFY`, so one set of values serves both, and a synced chunk carries its page id so the Executive can open the live page with such a tool. A Confluence token is a user's, and what the sync reads becomes retrievable by everyone who talks to the Executive, so the space list is required and pages with a read restriction (their own or a parent's) are skipped unless `CONFLUENCE_SYNC_SKIP_RESTRICTED=false`. That keeps any existing member's token safe to use, without a dedicated licence. Manual runs: `openexecutive sync-confluence` / `purge-confluence`.\n\n### Your own Gmail or Outlook (Act as me)\n\nA second, separate mail connection, for **Act as me** (see **Act as Me (Delegation)**): a person's *own* mailbox, which `delegation.gmail.DelegateGmail` reads and drafts into over the Gmail REST API. It is never registered with the MCP gateway, so the model can reach it only through the typed `ghostwrite_email` handler — never by tool name through `call_tool`. The credential is one file per person in `DELEGATION_GOOGLE_CREDENTIALS_DIR` (default `company/delegation_google/`, `/data/delegation_google` in Docker), minted on a computer with a browser by `scripts/connect-own-gmail.py --email <you>` with the Executive's own OAuth client and the scopes `gmail.readonly` and `gmail.compose` — the same seed-it-yourself flow as the Executive's token, since the API serves no OAuth callback. It never lives in `WORKSPACE_MCP_CREDENTIALS_DIR`, where workspace-mcp picks a credential by address.\n\nEvery use re-checks with Google that the credential opens the mailbox on the person's People entry, and refuses outright when that address is the Executive's own: a shared mailbox would make everything the Executive writes look like theirs. Drafts are threaded (`In-Reply-To`, `References` of whole ids ending with the parent, `threadId`) and stamped `X-OE-Ghostwritten`; the only send is `drafts.send` of an existing draft by id, when the person taps Send on its reply card (`delegation.reply_send`). A Gmail rate limit, even as a 403, is an error to retry rather than a reconnect. The voice learner and the thread shown to the ghostwriter keep only each sender's own words through `email_poller.sender_new_text`, the poller's quoted-reply stripper made public. Setup status shows it as **Your own mailbox (Act as me)**.\n\nThe same file can instead hold an Outlook sign-in (`\"provider\": \"microsoft\"`), minted by `scripts/connect-own-outlook.py --email <you>` with a device-code sign-in on the Executive's own Entra app (`MS365_MCP_CLIENT_ID`, `MS365_MCP_TENANT_ID`, default `common`) and the delegated scopes `User.Read`, `Mail.ReadWrite`, `Mail.Send` and `offline_access`. `delegation.gmail.gmail_for` then hands back `delegation.outlook.DelegateOutlook`, which talks to Microsoft Graph directly — never the Microsoft 365 MCP server the Executive's own Outlook uses — with the same methods and the same one send: Graph's `send` on an existing draft, with no body. Microsoft rotates the refresh token on use and the new one is written back into the file. A sender counts as verified only from the topmost `Authentication-Results`, the one Exchange wrote, and only mail in their Sent Items counts as theirs.",
+ "markdown": "External channels connect callers to the Executive. Each inbound adapter resolves the sender to a Person — the chat adapters via `find_person_by_{slack,discord,telegram}_id`, the email poller via `people.identity.resolve_email_sender` (a primary address or an alias, then the company-domain rule; see Org) — and threads `person_id` into `Executive.chat()`. A sender who resolves to nobody is no longer simply dropped: see **New senders (roster requests)** below.\n\n### Inbound channels\n\n- **Slack** (`integrations.slack_bot`) — async Bolt app (`AsyncApp`) embedded in the lifespan, hearing Slack one of two ways: socket mode (`AsyncSocketModeHandler`, with `SLACK_APP_TOKEN`) or, with only `SLACK_SIGNING_SECRET`, the Events API over HTTP at `POST /webhook/slack` (`integrations.slack_events`), where the route checks each request's signature and age (and a 1 MB cap) before Bolt sees it, so unsigned traffic never reaches Bolt's logging, and Bolt acknowledges at once while the listener runs. Either way a repeated event id is dropped (Slack resends what it thinks went unacknowledged), as is an event delivered for another workspace. Responds to `@mention`s, DMs, and thread continuations, with per-conversation session memory (see **Conversation state** below). Text only: there is no interactivity endpoint and no Block Kit, so nothing Slack sends carries a button.\n- **Discord** (`integrations.discord_bot`) — discord.py gateway embedded in the lifespan; per-thread auto-titling, session-scoped memory, author attribution, configurable `@mention` thread promotion.\n- **Telegram** — webhook gating + per-chat history. Replies (and every outbound Telegram message) go out as Telegram's HTML subset (`utils.markdown_telegram`: markdown-it-py with raw HTML and images off, text escaped), so the Executive's Markdown shows as formatting; links show as their text and target in plain text and never become tags, so quoted text can't hide where one goes; a chunk Telegram refuses is sent again as plain text.\n- **Google Chat** — webhook + service-account auth. It resolves no approval gates and has no roster gate (the webhook authenticates Google, not the sender), and it never populates the session's trusted alert ids, so `ack_alert`'s trust check — which now runs on every session rather than only those declaring an `origin_channel` — refuses there by default, and its `<channel>` block says plainly that a sign-off cannot be answered from it.\n- **Email** (`integrations.email_poller`) — MCP poll loop over the Executive's own mailbox — Gmail via `workspace-mcp` or Outlook via `ms-365-mcp-server`, chosen by `EMAIL_PROVIDER` (see *Workspace providers* below); it starts only once the MCP gateway is up, never surfaces Reply-To redirects, filters auto-replies. Its sessions carry `origin_channel=\"email\"`: the untrusted-content policy (`orchestrator.content_trust`) reads that tag to tell mail from the principal's own surfaces — left empty, as it once was, every stranger's body reached the decision extractor as the principal's words. The mail goes into the turn inside an `<untrusted_content source=\"email\" author=\"<sender>\" author_verified=\"no\">` block that says it carries none of the principal's authority, with the framing line and any `[POLICY]` notice outside it; mail from the principal's primary address that Gmail authenticated (`dmarc=pass`) is their own words — but only their new text goes in unwrapped: when the mail quotes or forwards another message (`_new_text_lines` cut something), the message as received follows in an `<untrusted_content source=\"email_quoted\">` block, since a stranger's text they replied to or passed on is not theirs. While the Executive is paused (`scheduler.pause`) it skips each poll and leaves the inbox unread, so held mail is processed on the first poll after resume; the chat adapters above are not paused. The Executive sees the whole message; peer memory records only `_email_memory_text` — the subject, the sender's new text (cut at the first quoted or forwarded message) and the attachment filenames — parsed from the rendered message with any trailing `--- REPLY ---` block dropped first, without backtracking-prone regexes, since the text is sender-controlled, and only for a rostered sender, the only kind with a peer to record into. The poller's `integration_inbound` row keeps the mail's text for the audit log, the way the reply's `chat_turn` row keeps the Executive's answer: `full.message` is the sender's new text (what web chat keeps for a turn), `full.attachments` the attachment filenames, and `details.preview` the first 120 characters of the new text, so /audit shows what was asked next to what was answered (the summary still reads `Inbound email from <sender>: <subject>`). The whole body with its quoted chain (`full.body`) is kept only on a row private to the principal: the chain is other people's mail — a contact's, on the principal's reply to one — and every signed-in user reads a public row. Each field is cut by its escaped JSON length (20,000 characters of text, 8,000 of attachment names), the measure the logger's 64 KB cap uses, so a long mail in a non-Latin script loses its tail rather than the row's structure. Outbound, the principal briefs and the client-rotation digest are emailed through the same gateway with the configured backend's send tool (`google_workspace__send_gmail_message` or `microsoft_365__send-mail`, via `workspace.registry.send_from_executive`, from the Executive's own mailbox to the principal's address, subject `<brief title> — <date>`, the date in the user's zone) as formatted HTML — rendered from the brief's Markdown with raw HTML and images off, so text quoted from outside shows as text and opening the email loads nothing remote. Email is first when the principal prefers it, and otherwise the backup after every chat channel, tried only when Slack, Discord and Telegram are missing or fail; an in-band error from the tool counts as a failed send and the next channel is tried. An emailed brief is not a DM, so the per-Person outbound guard below does not apply — the gateway's roster egress gate does.\n\n Two senders get their own framing. Mail from one of the principal's **contacts** (a `kind=\"contact\"` People row — see Org) replaces the non-roster `[POLICY]` notice with a contact one: it names them (name and role/company), says not to reply unless the principal asks and not to tell anyone else, and asks for a summary for the principal instead. Both this turn and the forwarded-by-principal turn below are marked `Session.private_to_principal`: they can email, DM or invite only the principal, an alert they raise is private to the principal, and they are not offered the tools that reach anyone else or start work outside the turn (`schedule_tools.PRIVATE_TURN_WITHHELD_TOOLS`: broadcasts, meeting bookings, workflows, follow-ups, watches, artifacts, goal updates, skill drafts, connecting a new MCP server), and of the MCP tools only Google Workspace reads and the gated Gmail send (see MCP below). Every audit row about such mail is private to the principal (see Audit Log), the poller's two `integration_inbound` rows included (`_private_to_principal_mail`), so the audit log's other readers see none of it. The poller holds `audit.private_rows` for the whole handling of such a mail, because some rows are written before the turn binds its private session: the knowledge retrieval quotes the mail's From line, subject and opening. The rows themselves read exactly like an outside sender's: the `integration_inbound` row is the non-roster one (`accepted_non_roster`), and the notice opens with the non-roster notice's exact sentence. The poller handles each message with no session bound, so one message's private turn never marks the next message's rows. The turn stays unattributed (`person_id` None — no peer memory, no co-presence), and the gateway refuses mail to a contact on a poller turn anyway. Mail **forwarded by the principal** — the sender resolves to the principal and `_new_text_lines` reports a forwarded message — is wrapped in a `<forwarded_by_principal>` block: summarise it, draft a reply the principal could send (as text in the reply, never sent or saved as a Gmail draft), note any commitment or date, and offer to add the original sender as a contact (confirmed from the web app or the principal's Slack / Discord, since an ordinary email turn cannot change the People list — the one email that can is the principal's tokened reply to a roster-request email, which is parsed before any model turn; see below). The reply still goes to the principal only. The From header is unauthenticated, so this framing grants nothing: a spoofed \"forward\" just gets the real principal a summary.\n\nSlack, Discord, and Telegram are roster-gated on inbound: an unrostered sender's message never reaches the Executive (`details.outcome=\"rejected_unknown_sender\"`). The gates look up **team members only** — a contact's Slack, Discord or Telegram id is treated as unknown, so a contact can never talk to the bot (nor count as a co-present peer). What happens next is the roster request below — except for a contact, a thread continuation, a Telegram group or an unverified Telegram webhook, which stay silent. Email is not inbound-gated — non-roster mail flows through with a `[POLICY]` notice so triage can classify it. Slack/Discord thread continuations pass the shared `integrations.response_gate` (utility-fast model, biased toward replying). Slack only continues a thread it has replied in: it reads the thread with `conversations_replies` (5 s bound) and looks for its own user id, which `auth_test` resolves at startup. If that lookup fails, mentions and DMs still work and thread continuations stop until a retry succeeds (`slack_bot._resolve_bot_user_id`, at most once a minute, from the channel `message` listener). When Slack doesn't return the thread in time, the thread's stored history (`slack:thread:{channel}:{thread_ts}`, written on every turn the bot answers there) stands in: with history, a rostered sender gets an ephemeral \"I couldn't read this thread just now… @-mention me\" and the drop is audited (`outcome=dropped_thread_unreadable`). A thread the bot never answered in stays silent. A message the gate skips is audited (`outcome=skipped_gate`, with the gate's reason) and gets an :eyes: reaction (`reactions_add`, which needs the bot's `reactions:write` scope; without it the reaction is skipped), so a sender the gate misjudged can see the message was read and @-mention the bot.\n\n**No link previews on replies.** The Executive's replies go out with previews off — Telegram `sendMessage` carries `link_preview_options.is_disabled`, Slack's reply `say` passes `unfurl_links=False` / `unfurl_media=False`, and every Discord reply send passes `suppress_embeds=True`. The direct messages the Executive starts are the same: `send_slack_dm` posts with `unfurl_links=False` / `unfurl_media=False`, and Discord's `send_dm` sets the `SUPPRESS_EMBEDS` flag on each chunk, so a reminder or a brief sent by DM carries no preview either. Otherwise the chat service fetches a link in the text on its own, so a reply that quotes someone's mail (Act as me reads it) could carry data to a sender's server with no click. The web chat does the same for images: `Message.tsx` loads only an image the app itself serves (`lib/markdownImages.loadsInline`) and shows any other as a link.\n\n### Alert triage of chat messages\n\nEvery chat-app message that passes the roster gate (Telegram, Slack, Discord, Google Chat) is also handed to alert triage through `alerts.pipeline.schedule_chat_message_evaluation`. The sender is named by their People record (`full_name`), not by the name they typed into the chat app, so a person whose account there is called something else is never read as a stranger; the chat app's own name (or raw user id) is used only when nobody on the roster matched. The principal's own messages to their Executive are a conversation, not news for their brief, so they are not triaged at all. Triage sees the sender as `sender:` and the room as `channel:`.\n\n### New senders (roster requests)\n\nSomeone who writes in and is not on the People list — and is not one of the principal's contacts — is **held, acknowledged and asked about** instead of dropped (`integrations.roster_intake.intake` over the `people.roster_requests` ledger; lifecycle in Org). Each adapter calls it after its gate:\n\n- **Slack** — a DM or an `@mention` (not a thread continuation: the bot was not addressed there). The name and profile email come from `users_info` (5 s timeout). The acknowledgement goes in the DM, or as an **ephemeral** message for a mention, so nobody else in the channel learns who is and is not on the roster. The held event keeps only a fixed set of keys (text, user, channel, ts, thread, files).\n- **Discord** — a DM or a mention (not a thread the bot merely follows). The acknowledgement goes in the DM, or by DM for a channel mention; `/ask` answers the interaction with an ephemeral acknowledgement first (Discord allows 3 s), then holds the prompt. Its old \"ask the admin to add you\" reply is gone.\n- **Telegram** — only a **private** chat, and only when `TELEGRAM_WEBHOOK_SECRET` is set and valid: Telegram identifies the chat rather than the sender, and without the secret anyone could forge an update. A group, or an unverified webhook, stays silent as before. Attachment file ids are held and downloaded only on replay.\n- **Email** — non-roster mail that is not private to the principal (so never a contact's) and does not look machine-sent (`looks_automated`: a no-reply / notifications / billing-style local part, a `List-Id` / `List-Unsubscribe` / `Auto-Submitted` / bulk `Precedence` header, or a DMARC failure). The mail is **still triaged** by the Executive on this pass; its `[POLICY]` notice now says the sender was already told their message is waiting and the principal was asked who they are, so the model raises no second alert just to add them, and that the email will come back once they are added.\n\nEvery path skips a contact first (`find_person_by_*_id(..., include_contacts=True)`, or the email poller's private-mail rule): a contact has no chat access and gets no request and no acknowledgement, and the team-visible `rejected_unknown_sender` / `accepted_non_roster` rows read the same for a contact as for a stranger.\n\n**The acknowledgement** is one fixed text (`roster_intake.ACK_TEXT` — \"your message arrived … waiting for confirmation\"), never the model and never an echo of what they wrote, so a forged sender gains nothing by triggering it. It is sent at most once per sender per `ROSTER_ACK_WINDOW_DAYS` (7) and at most `ROSTER_ACK_DAILY_CAP` (20) a day across all senders (`claim_ack`, one transaction; a failed send releases its claim). An email acknowledgement is the one send the MCP gateway lets reach an address off the roster: `send_email_ack` sets `mcp_gateway.roster_ack_grant`, a one-shot ContextVar grant that admits a single `send_gmail_message` with exactly the granted recipient (a plain string, no list), subject and body and no other key, then is spent (audited `integration_outbound`, `kind: roster_ack`, private). An email sender is acknowledged only when Gmail authenticated them: once `claim_ack` succeeds, the poller's `send_ack` re-reads the mail as raw MIME and needs Gmail's own topmost `Authentication-Results` reporting `dmarc=pass` for the From domain (`fact_confirmation.sender_authenticated`, fail-closed). The printed headers the poller reads never carry that header, so without the second read a forged From would draw the acknowledgement to whoever it names — backscatter from the Executive's mailbox. An unauthenticated sender makes `send_ack` raise `roster_intake.AckWithheld`: the claim is released, and they are held silently. `authenticated_by_gmail` strips quoted strings and comments before reading the verdict — Gmail copies the envelope sender and DKIM tags into its own stamp, and a quoted local part could otherwise carry a fake `dmarc=pass` — and needs exactly one `dmarc=` verdict reading exactly `dmarc=pass header.from=<domain>` (Gmail may follow it with its own `dara=` resinfo, so its position is not checked). Residual: when Gmail writes no verdict at all (a From domain without DMARC) and echoes an unsanitised `;dmarc=pass header.from=…` elsewhere in its stamp, that echo would read as the verdict; bounded by the acknowledgement caps. A withheld acknowledgement is audited `roster_ack_withheld`. `intake` acknowledges before surfacing the card and telling the principal, and both follow the request's `ack_sent_at`: a withheld, failed or capped acknowledgement reads \"haven't been told anything yet\" on the Today card, in the principal's prompt and in the email `[POLICY]` notice.\n\n**Telling the principal.** A new request puts a private card on the principal's `/today` (see Today) and pushes a prompt on their delivery plan — preferred chat DM, then the others, then email — for a chat sender or an email sender on the company's own domain; an outside email sender waits on the card, as does everything past `ROSTER_REQUEST_DAILY_CAP` (30) pushes a day (the same cap bounds how many requests open a day; a declined sender is not raised again for 30 days). A chat prompt can be answered in place (\"that's Annamarie, add her to the team\"). An email prompt carries a one-time token (`RR-` + 20 base32 characters, stored hashed); the poller checks every inbound mail for one **before** triage (`try_email_roster_answer`): the reply must come from the principal's primary address exactly (not an alias, not the company-domain rule), carry the token of a pending request, not be an automatic reply, and not fail DMARC where the header is visible; and is parsed by `parse_decision`'s `roster_identity` shape — anything else falls through to the ordinary path. The token is hidden (`RR-[hidden]`) in every other Google Workspace result the gateway returns — the confirmation sits in the Executive's Sent folder, and any turn could otherwise read it — except the poller's own fetch of an inbound message (`mcp_gateway.reveal_roster_tokens`); a token-bearing mail that is not an answer is scrubbed before triage.\n\n**Replay.** Once the principal answers (web card, chat tool, email reply) — or a People-page / `upsert_person` write makes the sender match someone (`after_roster_write`) — the held messages are replayed through the adapter that received them (`register_replayer`: Slack re-runs the held event, Discord and Telegram their held text, email re-fetches the message from Gmail and handles it as mail from a known sender), oldest first, now attributed to the new Person. A teammate gets every held message; a new contact only held email (it takes the private contact path; held chat is dropped); a decline drops everything. Replays run in a **fresh `contextvars.Context`**: the answer may have come from the principal's own verified turn, and a stranger's message must never be read with the principal's session — and their contacts — still bound.\n\n### Standing facts by email (confirm by reply)\n\nThe principal can ask for a standing fact, a retirement or a company-profile edit by email, and the fact tools check it exactly as in chat (the own-words gates run on `_email_memory_text`, the sender's new text, not the quoted chain or attachments). An email's From line proves nothing, though, and a standing fact is read by every later prompt as the principal's own account. So an emailed change is **held, never applied**, until the principal's own mailbox confirms it (`integrations.fact_confirmation`, `memory.facts.fact_confirmations`):\n\n- **Who may ask.** The poller stamps every email turn with `Session.email_from` (the bare lowercased From address) and `email_authenticated`. The text the poller reads never carries `Authentication-Results` (workspace-mcp prints a fixed set of headers), so for mail claiming the principal's primary address it reads the message again as raw MIME (`fact_confirmation.sender_authenticated`) and requires Gmail's own stamp: the **topmost** `Authentication-Results` header (Gmail adds it on receipt, above anything the sender wrote), authserv-id `mx.google.com`, reporting `dmarc=pass` for the From address's domain, with the raw From matching (`authenticated_by_gmail`). It fails closed: no header, `dmarc=none`, a temporary error or a failed read all leave the turn unauthenticated. Mail from a teammate other than the principal is read the same way, before its inbound audit row is written: authenticated, the whole handling of it is theirs alone in the audit log (`rows_for_person`, see Audit) and the turn reads their own mail as them (`Session.email_sender_authenticated`); unauthenticated, it keeps the ordinary rule. Mail from anyone else costs no second read and is never authenticated. `fact_tools._gate` admits an email turn only from the principal's **primary** address, exactly (no alias, no company-domain rule), authenticated, and not private to the principal (mail they forwarded stays refused, as do contacts, teammates and unknown senders).\n- **The hold.** The tool returns `awaiting_confirmation` (no action chip, and the model is told to say nothing has changed). `request_confirmation` stores the change with the SHA-256 of a fresh one-time token (`FC-` + 20 base32 characters, 100 bits; 7-day expiry; at most 10 waiting at once) and emails the token to the principal's roster address, outside any turn's outbound context. A profile edit is dry-run first, so one that cannot apply is reported in the turn, before any email goes out. If the email fails, nothing is held.\n- **Keeping the token from any model.** The tool result never carries it, and `mcp_gateway.hide_roster_tokens` now scrubs `FC-` tokens as well as `RR-` ones (`FC-[hidden]`) from every Google Workspace result except the poller's own fetch of an inbound message. Act as me's reads of the principal's own Gmail, where the confirmation email lands, hide them too (`delegation.gmail.parse_message` and the thread search), so the drafting model never sees a live token either.\n- **The reply.** `_handle_one_email` hands every message to `try_email_fact_confirmation` right after the roster-answer check, before any model turn. It acts only on a reply from the principal's primary address carrying the token of a **pending** confirmation. A made-up, spent, cancelled or expired reference matches nothing and falls through to the ordinary path with the token hidden, so no old or forged token can make the Executive email the principal. A reply with a live token is read again as raw MIME (`read_raw`) and must pass the same Gmail DMARC check as the request; one that doesn't (or whose raw read fails) is refused, the change keeps waiting, and the principal is told. Then an **automatic reply** is consumed quietly — no answer, no model turn, the change keeps waiting: the printed headers catch a Gmail vacation reply (`Precedence: bulk`), and the raw headers catch what they don't show (`automatic_reply`: `Auto-Submitted` other than `no`, `X-Auto-Response-Suppress` and similar, `Precedence: auto_reply`, an out-of-office subject), since an Exchange out-of-office echoes the token from the principal's own address and may well say \"I'll confirm on my return\". The raw message starts at the tool's exact separator line (`\\n\\n--- RAW MIME ---\\n`), so a sender-written Subject imitating it cannot supply the headers. CONFIRM (or yes / approve) applies the change through `fact_tools.apply_confirmed`, the same apply step chat uses. CANCEL (or no) drops it, and an unclear reply leaves it waiting. The principal gets a one-line result either way. A compare-and-set on the row means two replies can apply it only once.\n- **Audit.** Private `fact_confirmation` rows record held / applied / cancelled / refused. The applied change also writes the fact tools' usual private rows, with `source_channel=\"email\"` as its provenance.\n\n### Checking a channel is set up\n\nSettings → Setup status (`GET /setup/status`, `api/setup_checks.py`) gives each channel a light and a plain-language fix. It tests the credentials with one free read-only call — Slack `auth.test`; Telegram `getMe`, then `getWebhookInfo` (is a webhook registered, is it this app's `/webhook/telegram`, did a delivery fail in the past day); Discord `users/@me`, only while the bot is still connecting — and reads what the lifespan left on `app.state`: the Slack socket's `is_connected()`, the Discord bot's `is_ready()` together with its gateway socket being open (discord.py stays ready while it reconnects after a drop), or the exception its task stopped with (`LoginFailure`, `PrivilegedIntentsRequired`), and the MCP gateway. A listening Slack is amber while the listener hasn't resolved its own user id (`slack_bot.bot_user_id()`), since thread continuations are off until it does. Because inbound is roster-gated, a connected Slack, Discord or Telegram is amber while nobody on the roster has its id (`slack_user_id`, `discord_user_id`, `telegram_chat_id`), or when its latest `integration_inbound` row is a `Rejected: …` from the past day whose sender still isn't on the roster — the page shows the sender's id, never message text. Local login with no `TELEGRAM_WEBHOOK_SECRET` is red for Telegram, since the local-login gate turns its webhook away. Google Chat and Gmail are checked from configuration: a numeric project number and a readable service-account key; the OAuth client (or service-account key) the workspace-mcp child signs in with. Gmail is also amber when startup priming could not find every pinned Google tool (`MCPGateway.prime_pinned_tools`, whose result the gateway keeps on `pinned_tools`): the row names a few of the missing tools and counts the rest, and points at the `MCPGateway` log lines. It is amber too while that probe has not finished, and when the probe itself failed — the one check that reads green is the one that ran and found everything, so a crashed probe is never silently fine. The row speaks for the whole `google_workspace` server, so a missing Docs or Drive tool shows up under Email (Gmail).\n\nA channel setting written `KEY= # note` in a `.env` file arrives as the note's text, because python-dotenv strips only a comment that follows a value. `Settings` reads such a value as unset for the Slack, Telegram, Discord and Google Chat credentials, so the line no longer switches a channel on with its own comment as the token. `TELEGRAM_WEBHOOK_SECRET` is the exception, because read as unset a junk secret would turn the webhook's check off. It is kept instead, and since Telegram can only ever send 1–256 of `A-Z a-z 0-9 _ -` (what `setWebhook` accepts), a secret outside that set — a leftover comment, a pasted `<value from Step 2>` — could only be matched by someone who guessed it: while it is set the webhook refuses every update, neither the local-login gate nor the roster tools count it as proof (`Settings.telegram_webhook_secret_valid`), and Setup status shows it red (`.env.example` shipped four such lines; the notes now sit on their own lines).\n\n### Conversation state\n\nEvery adapter keys a deterministic `session_id` off the conversation surface and replays that session's stored turns into `Session.conversation_history`, so a follow-up (\"yes, go ahead\") lands with the question still in context. Slack's ids are `slack:dm:{user}`, `slack:thread:{channel}:{thread_ts}`, and `slack:channel:{channel}:{user}` for a mention that starts a thread; because the bot's reply roots a Slack thread at that message's `ts`, the turn is double-written to that thread's session too, so the first continuation inside it is not cold. A reply inside a thread THAT PERSON ROOTED also offers the parent `slack:channel:{channel}:{user}` id when matching an approval gate, since a gate raised on the mention recorded that session while the answer arrives under the thread's. The root-author check is load-bearing: widening on \"threaded and not a DM\" alone offered one person's channel session inside any thread in that channel, so a remark they made to a colleague could resolve their open gate and post the run's title into that colleague's thread. Discord has the same shape — its auto-thread router promotes a mention reply into a bot-owned thread — and the same alias. An @-mention inside a DM is handled once, by the DM path: Slack fires both events, and taking both forked the conversation across two session ids. The same id is the audit grouping key and the scope passed to `format_for_prompt`. A per-session `asyncio.Lock` serializes turns within one conversation — the inflight semaphore caps fan-out across the workspace, which does nothing to stop two messages in the same thread from both reading history before either writes it. Persistence happens after the reply is sent and its failure is logged, never surfaced: the user already has their answer.\n\n**Conversations within a chat app.** A Telegram chat, a Slack or Discord DM and a person's @mentions in one channel (`slack:channel:…`, `discord:channel:…`) are each one endless stream, so they are cut into conversations (`memory.conversations`, shared helpers in `integrations/conversations.py`). Each adapter computes the stream's base id as before and asks `conversation_for` which conversation a message belongs to: the newest, unless the stream has been quiet for two hours (`CONVERSATION_GAP`), when the next one starts. The first conversation keeps the bare base id; later ones add `~2`, `~3`, … (`memory.conversation_ids`). A whole message of `/new`, \"new chat\" or \"new conversation\" starts a fresh one instead of a turn and is answered with the name of the one it ended; the new row has no messages and stays off the Chats page until its first turn. After a conversation's first exchange `title_after_turn` names it in the background with `utils.session_title.generate_session_title` (the first words of the message if that fails), as a web chat's first turn does. Threads (`slack:thread:…`, `discord:thread:…`), email and every other id are left whole. What has to see the whole stream compares base ids: the inbound resolver, so an approval asked before a cut is answered after it, and the session-scoped decisions and advice `format_for_prompt` renders. At boot `split_existing_streams` cuts history saved before this by the same rule, once (`app_migrations`), titling each part from its first message and carrying a read-mail lock to every part. An adapter installed outside the package adds its streams with `conversation_ids.register_rolling` and cuts its own old history once, with its own migration name and an `only` filter for its base ids, because the built-in sweep may already have run without it.\n\nThis was absent on Slack until #136. Every message built a fresh `Session` and minted a per-*message* id (`thread_ts` falls back to the message's own `ts`, so a DM never repeated one), which meant a multi-turn confirmation could not work: the Executive asked whether to send a briefing to Slack, the user said yes, and the next turn had no record of the question.\n\n### What the model is told about the channel\n\nChat adapters pass a `<channel>` block (`integrations.channel_context`) in the **user turn** — never a cached system block. It names the surface, states that the person cannot see the app's cards or buttons, and draws the line the assistant kept crossing: briefing proposals and alerts are actioned on the briefing page; a proposal that books something can only be approved there; a workflow sign-off the Executive is explicitly waiting on *can* be answered in the conversation. The principal's DMs on Slack, Discord and Telegram also receive the open-alert digest (`briefing.context.format_open_alerts_for_prompt`) as the `<briefing>` block, gated to DMs because the board is company-wide and a shared channel must not pull it in. That digest is the ONLY trusted source of an `alert_id` for `ack_alert`. The briefing page's `[Discuss mode — alert_id=N]` primer is a pointer to which item is under discussion, not authority on its own: the server accepts that id only if it is also on the live board it derived itself. An id appearing anywhere else, including inside an alert's own body, is not trusted: alerts are minted from inbound email and chat, so their bodies are attacker-controlled. Two mechanisms enforce that rather than one prompt rule. Every field the digest interpolates is collapsed to a single line, because a newline in an attacker-written headline, suggested action, tag or review note would otherwise forge a whole extra `[N]` line — a fabricated instruction to clear somebody else's alert. And the live board the server derived is recorded on the session, so `ack_alert` refuses any id outside it — on EVERY session, web included; the web previously had no check at all. The recorded set is the whole live board (the cap `/today` renders as cards), not just the rows the digest printed, so a card past the digest's 30-row render cap is still ackable rather than sending the principal back to the page. Be precise about what this buys: it stops an id that is invented, closed, snoozed or expired, including one quoted out of an attacker-written alert body. It does NOT stop the model being argued into acking the wrong LIVE card — one alert's body can still make a case about another id that is genuinely open. Scoping trust to the single item under discussion would need the primer id carried server-side, which it is not.\n\n### Handler failures\n\nAn adapter's catch-all replies \"I encountered an error processing your request.\" On Slack that used to be the whole story — one uncorrelated log line, no audit row, and an unguarded error-send that could itself escape into Bolt. It now logs with the session, sender, channel, thread and mode, and writes an `integration_inbound` row with `details.outcome=\"handler_error\"`, joining the inbound row under the same `session_id`. Upstream, a single crashing tool handler no longer takes the turn down at all: the orchestrator's tool gathers use `return_exceptions=True`, hand the model an error `tool_result`, and audit it with `ok=false` (cancellation is re-raised, not swallowed).\n\n### Attachments\n\n`integrations.attachments` handles web chat, Discord, Telegram, Slack and Google Chat. Images become Anthropic vision blocks; documents (PDF/DOCX/TXT/MD/CSV) are inlined into the turn that carried them — inside an `<untrusted_content source=\"attachment\" author=\"<name>\">` block (`content_trust.wrap_untrusted`, whose closing tag the text cannot forge), under an `[Attached: <name>]` label, which the open-loop pass reads as \"not the speaker's words\" and the decision extractor strips before it reads the turn — and indexed into the ChromaDB `inbound_attachments` collection. 20 MB/file cap; web chat uses `POST /chat/upload`.\n\n**Scanned PDFs** are read, not reported empty. A PDF with no text layer goes through `knowledge.pdf_reader`: with `PDF_PROVIDER_READING=true` (opt-in, off by default) the deployment's own model reads it through its provider's PDF support (Anthropic natively, OpenRouter's `file-parser`, a local server only with `LOCAL_PDF_INPUT`); otherwise, and as the fallback, local OCR (pypdfium2 + RapidOCR, offline, no key) — so every deployment converts them. The label then reads `(converted from scanned pages)`, and an unreadable one says why. The index takes the text already extracted for the turn, so a scan is converted once. A file a channel delivered (everything here except `POST /chat/upload`, where the signed-in user sends it) converts at most `PDF_INBOUND_MAX_PAGES` (30) pages and draws on a rolling `PDF_INBOUND_PAGES_PER_HOUR` (300) budget shared by all senders, so sending scans cannot run up unbounded model spend or CPU.\n\nPer channel:\n- **Slack** — a `file_share` message (previously dropped outright for its `subtype`) is handled like any message; files download only after the roster gate and the response gate, with the bot token sent only to `https://files.slack.com/` (needs the `files:read` scope). History and peer memory record the words plus `(Attached files: …)`, never a document's text.\n- **Google Chat** — uploaded attachments download through `media.download` with the reply credentials (`chat.bot`), checked against the 20 MB cap per chunk, 5 per message; an attachment-only message is no longer dropped. The webhook authenticates Google, not the sender, so files are read only when the event's `user.email` is on the People roster (a team member); anyone else's files are named to the Executive, never downloaded. A Drive file shared into a message is left to the Drive tools, with its id.\n- **Email** — the poller (`integrations.email_attachments`) downloads up to 5 readable attachments of the message itself via `get_gmail_attachment_content` — workspace-mcp in stdio mode saves the file and answers `Saved to: <path>`, which must start its line, be the last such line and end in the answer's `Saved filename:` (the answer echoes the sender-chosen attachment name first, and a name carrying `Saved to: /…` must not redirect the read), then resolve inside the tool download folders — and appends their text to the turn (15k chars each). Only for a sender the principal knows (team, contact, principal); a stranger's files stay a list. It runs outside the model loop, so it works on private turns without adding the attachment tool to `PRIVATE_TURN_MCP_TOOLS`.\n- Anything else (a Drive PDF, a skipped attachment) — the Executive's `read_document` tool.\n\nIndexed chunks are keyed on `attachment:<bare filename>` and tagged `type=attachment`. An attachment's name is chosen by whoever sent the message and the name is the chunk-id namespace, so without the prefix an inbound `strategy-2026.md` would upsert over the curated company document of that name. The prefix separates inbound content from curated uploads and keeps provenance visible; it does not separate senders from one another.\n\nIsolation is the **collection**: `retriever.retrieve` never queries `inbound_attachments` at all, so an attachment cannot resurface as company knowledge in a later, unrelated turn. Attachment processing is roster-gated on Discord too — it previously ran before the gate, so an unrostered sender's file was downloaded and indexed before their message was dropped. These rows still have no operator surface: they are never written to `company/docs/`, so `GET /documents` does not list them.\n\nThis replaced a domain-based scheme that did not work. The chunks lived in `company_docs` under a domain outside the specialist set, on the premise that an unrecognised domain matched no filter. It matched every one — `ChromaDBStore.query` attaches a `where` clause only when a domain filter is present, and an Executive-level retrieval passes none — so attachments came back under “From your company documents”, top-ranked and indistinguishable from a curated upload. Legacy rows are migrated across on boot.\n\n### MCP gateway & outbound\n\nMCP runs in **both directions**. As a *client*, `orchestrator.mcp_gateway.MCPGateway` proxies MCP servers (Gmail, Calendar, Drive, Outlook mail and calendar — whatever `mcp_servers.json` configures) as Executive tools, with egress gating on Gmail recipients, Calendar attendees, Drive share grantees, and Microsoft 365 recipients/attendees against the roster (and a refusal of public / whole-domain Drive sharing). The Gmail gate also accepts only a fixed set of argument keys, so workspace-mcp's `reply_all` (recipients taken from the thread, never checked) and `forward_message_id` / `include_forwarded_attachments` (forwarding any inbox message) are refused; a reply is `thread_id` plus an explicit, roster-checked `to`. Three more refusals close paths around those gates: every Apps Script tool (code running as the Executive's account, outside the gates, and since workspace-mcp 1.29.0 schedulable to keep running); any argument that makes the server fetch a URL (`fileUrl` / `file_url` on the Drive and import tools, a Gmail attachment given as a `url`), since the fetch can carry data out in its query string before any recipient gate matters; and text Google would mail to someone the roster never checked (`rsvp_comment` to an event's organizer, an out-of-office `decline_message` to whoever invites the Executive). Every gate reads its arguments by normalized spelling: workspace-mcp 1.29.0 renames a camelCase key such as `Attendees`, `userGoogleEmail` or `rsvpComment` to the real parameter, so a gate matching the exact key would never see it. Arguments that are not a JSON object are refused for the same reason. Google-side fetches are refused too: any key ending in `url` / `uri` at any depth whose value has a URL scheme (a Slides image or page background, a Docs or Forms image; a document hyperlink and a meeting link are stored, not fetched, and pass, as does a Drive file id), (a value is a URL after the characters a URL parser strips are removed, and `//host` counts), and on every tool that can write cells — the Sheets tools and the Drive create / update / import tools, which turn CSV into a Sheet — any string anywhere holding a formula call that fetches (`IMPORTDATA`, `IMPORTXML`, `IMPORTHTML`, `IMPORTFEED`, `IMAGE`), JSON-encoded or mid-CSV included: a string argument that encodes JSON is read as the server decodes it (one too large to read is refused), text is also scanned with character references decoded and comments / CDATA removed, and an upload that would become a native Sheet may only be text `content` — a `base64_content` or server-side `file_path` upload there is refused, since an XLSX hides its formulas behind zip members and XML encodings; a PNG to Drive or a DOCX to Docs is untouched. Not covered, unverified: whether Drive's HTML / Markdown import fetches an `<img src>` when it converts `content` into a Doc. A `manage_event` call that names no attendees is sent with `send_updates=\"none\"`, so the event's existing guests are not emailed text the gate never checked. The allow-list (`_egress_sets`, built by `_roster_allow_set` as a `people.identity.RosterAllow`) is every team member's email and email aliases plus the Executive's own address, always, and — for a teammate — any address on one of the company's own domains whose local part (`+tag` dropped) matches theirs, the same rule inbound mail is matched by; a **contact's** email joins it only when `people_tools.contacts_reachable_now()` is true — a turn the principal started on a verified surface (web chat, their own Slack or Discord, a verified private Telegram chat), or inside `grant_contact_egress()` (the decisions approve route, when the approver is the principal). On any other turn — an inbound email, a teammate, the scheduler, a workflow, the alert review — a contact's address is refused exactly like a stranger's, so the refusal cannot reveal that it belongs to a contact (contacts are private to the principal). The single exception to the roster rule is the fixed roster-request acknowledgement above, under its one-shot `roster_ack_grant`. A turn about the principal's private mail (`Session.private_to_principal`) narrows the list to the principal's own address and the Executive's. Such a turn may call only the MCP tools in `schedule_tools.PRIVATE_TURN_MCP_TOOLS`: Google Workspace reads (`search_gmail_messages`, `get_gmail_message_content`, `get_events`, `list_calendars`, `query_freebusy`, and the Drive, Docs and Sheets reads such as `search_drive_files`, `get_drive_file_content`, `get_doc_content` and `read_sheet_values`) and the two Gmail tools this gate narrows to the principal (`send_gmail_message`, `draft_gmail_message`). It is an exact allow-list, not a server: workspace-mcp runs at the tier `WORKSPACE_MCP_TOOL_TIER` names, `complete` by default, where a Google Chat message to a space, a Docs, Sheets or Drive write into a file already shared with others, `manage_event` on a shared calendar (its gate checks attendees and pins notifications, nothing else) reaches people with no recipient check; and every other server has none at all (a Slack or Notion server, the default `fetch` server, whose URL can carry the turn's content, a server loaded at runtime). Its `search_tools` results show only the allowed tools (`workflows.tool_catalog.filter_search_results`; output it cannot parse reads as no match), and a `call_tool` naming anything else is refused with an error tool_result and an audit row private to the principal (`schedule_tools.private_turn_withholds`), a missing name included. `search_tools` and `call_tool` stay offered, so the tool list and its cached prefix do not change. The typed calendar tool (`_resolve_attendees`) and the chat DM tools (`message_person`, `send_telegram_message`, `send_discord_dm` and their person-id recovery) apply the same rule — a contact's id reads as an unknown one off the principal's turn — and a private turn may DM or invite only the principal (`PRIVATE_TURN_REFUSAL` otherwise); `message_person` never falls back to a \"Needs you\" alert for a contact, who cannot sign in to see it. The Google Workspace server (`workspace-mcp`) behind those Gmail/Calendar/Drive tools runs **co-located inside the API** as a stdio child of the gateway — fitting the single-tenant model (one install = one company = one exec Google account), so there's no separate service to operate. It's baked into the API image and launched by `docker/workspace-mcp-launch.sh`, which selects single-user OAuth (token persisted at `/data/google_credentials`) or service-account domain-wide delegation from `GWORKSPACE_AUTH_MODE`. The optional `WORKSPACE_MCP_TOOLS` (comma-separated service names, default `all`) limits which Google services it loads. The gateway forwards the Google secrets to the child via `_FORWARDED_ENV_VARS`. Every `google_workspace__*` call acts as the Executive's own account. workspace-mcp takes the acting account per call (`user_google_email`) and nothing used to check it, so under domain-wide delegation a model-chosen address could open anyone's mailbox in the domain; `mcp_gateway._check_acting_account` now refuses a value naming any other address — before every other gate, with an `integration_outbound_blocked` audit row. An absent value stays allowed (the server's own account).\n\n**What the model is told is connected.** `MCPGateway.start()` records the configured server names (`server_names`), and the Executive renders them with the channel settings into block 0's *Connected Systems* section (`prompts.connected_systems`; see the caching section). While Google is connected that section pins the Google tool names the Executive uses constantly (`GOOGLE_TOOL_MANIFEST`, narrowed by `connected_systems.google_tool_manifest` to the services and tier actually loaded) and tells it to call them with `call_tool` directly, with no `search_tools` round trip. The narrowing is the point: a name pinned for a service workspace-mcp never imported is not in the index, so the model calls it, receives extensible-mcp's “has not been discovered via search_tools” refusal, and has no way to square that with its own instructions. extensible-mcp only runs a tool one of the session's searches has returned, and remembers it for the session's life, so the gateway does that discovery itself: the lifespan starts `MCPGateway.prime_pinned_tools()` in the background once the gateway is up with `google_workspace` configured (an exact-name search per pinned tool, the way `workflows.tool_catalog.resolve` does it, over the same narrowed set the prompt advertises — `connected_systems.pinned_google_tools`, so priming never probes a service that was not loaded). A name neither the prompt nor priming can account for is logged once, naming both causes: excluded by `WORKSPACE_MCP_TOOLS` / `WORKSPACE_MCP_TOOL_TIER`, or renamed by workspace-mcp since the pin. A pinned call that still comes back with extensible-mcp's \"has not been discovered via search_tools\" refusal, for example before priming finished, is discovered and retried once. That happens after every gate has passed the same arguments, and only for pinned names: any other tool still needs the model's own search, and nothing here widens what a call may do, including the private-turn allow-list.\n\n**Microsoft 365** is the Outlook twin: `@softeria/ms-365-mcp-server` (a Node package, pinned) is installed in a throwaway Node build stage and only the `node` binary plus the package are copied into the API image, launched by `docker/ms365-mcp-launch.sh` as the `microsoft_365` entry with an **explicit tool allow-list** (`--enabled-tools`, an anchored regex) rather than the upstream `mail,calendar` preset — that preset also registers inbox forwarding rules, mailbox settings (external auto-reply), calendar sharing and event forwarding, which can address people the gateway never sees. Exactly the Outlook mail + calendar tools and the OneDrive tools are exposed (`microsoft_365__send-mail`, `microsoft_365__create-calendar-event`, `microsoft_365__search-onedrive-files`, …; hyphenated Graph endpoint aliases, proxied verbatim), and every admitted tool is either roster-gated or carries no recipient; the gate also covers the excluded writers in case an operator widens the list. Both servers can be configured at once. Auth is the MSAL device-code flow against the operator's own Entra app registration (`MS365_MCP_CLIENT_ID` is required; the server's built-in public client is refused); the refresh token persists at `/data/ms365_credentials` and is seeded once with `ms365-mcp-launch.sh --login` from inside the container — the server has a native device-code CLI, so there is no mint script and nothing to copy. The launcher scrubs the literal `$VAR` placeholders extensible-mcp leaves for unset variables (otherwise MSAL would be handed `$MS365_MCP_TENANT_ID` as an authority), and `MS365_MCP_EXPECTED_USERNAME` pins the bound mailbox. The egress gate for this family walks Graph's nested shapes (`body.Message.toRecipients[].emailAddress.address`, `body.attendees[]…`) — every email-shaped string outside the free-text fields must be on the roster, control characters and non-ASCII addresses are refused, unforeseen keys are still scanned (fail-closed) — and covers the whole write surface: send, reply, reply-all, forward, every draft tool, `update-mail-message` (it can PATCH recipients onto a draft) and create-/update-calendar-event; reads and `delete-calendar-event` (the only cancel path) pass through. Graph's reply, reply-all and send-draft actions carry no recipient in their arguments (the recipient is whoever the referenced `messageId` names), so for those the gate first reads the referenced message through the server and roster-checks the recipients Graph will derive (`replyTo` else `from` for a reply; `replyTo` and `from` together, plus the original to/cc, for reply-all; the draft's own to/cc/bcc for send-draft), refusing on any lookup failure; RSVP and cancel actions get the same treatment through an event lookup, because their comment is emailed to the organizer or every attendee; `download-bytes` is pinned to mail attachment bytes and OneDrive file content (`/drives/{id}/items/{id}/content`, optionally Graph's `?format=pdf` rendering), and a OneDrive result comes back as the file's extracted text rather than base64 — the attachment pin is what makes the poller's promise that an unrostered sender cannot be auto-replied to hold on Outlook as it does on Gmail; and the rendered inbound text quotes any forged `--- REPLY ---` line inside a body so the persona only ever reads the real block. Every mail write's `from`, `sender` and `replyTo` may name only the Executive's own address (the Outlook twin of the Gmail acting-account pin), so the model cannot send as a rostered colleague's mailbox or steer replies elsewhere; and `move-mail-message` into Deleted Items, Junk Email or Recoverable Items is refused, so a move cannot trash mail ungated. The server registers its account tools whatever its tool allow-list says, so the gateway itself refuses `login`, `logout`, `select-account` and `remove-account` (audited like any other refusal) and drops them from `search_tools` results — a chat turn cannot sign the mailbox out, switch accounts or start a sign-in; setup is only the launcher's `--login`, and the read-only `verify-login` / `list-accounts` stay callable. The example config's hyphenated patterns still deny `delete-mail-*`, `logout`, `remove-account` and `download-bytes-to-file` as defense in depth. A successful `send-mail` records the same outbound-context linkage Gmail does.\n\n**OneDrive in chat** (opt-in with `MS365_MCP_ONEDRIVE=true` and a fresh `--login`, since the server asks for its tools' permissions on every token refresh and a sign-in without file consent would otherwise fail mail and calendar too) matches what the Executive can do on Google Drive: search, list a folder, read a file, upload and edit, create folders, move, rename, copy and share; there is no delete tool. The launcher adds `Files.Read.All` at sign-in (`--extra-scopes`), so folders shared *with* the Executive can be read too; writes (upload, new folder, move, copy) land only in the Executive's own drives, checked against `list-drives` at call time, so a planted instruction cannot save company files into a stranger's shared folder. Sharing is gated like Drive sharing: `share-drive-item` invitees must be People on the roster and named by email (an `objectId` or `alias` grantee is refused, since nothing can check it), no `owner` role, and `create-drive-item-share-link` must be scoped to specific people (`scope: \"users\"`): an anonymous or whole-organization link, or one with no scope (the tenant default), is refused. An invite must require sign-in and grant only view or edit, a link must be a view or edit link, and `search-onedrive-files` takes plain words only, since its query goes into the request path unencoded.\n\nAs a *server*, `openexecutive.mcp_server` exposes OE's own company context and specialist council to external MCP clients at `/mcp` — the inverse direction; see the dedicated **MCP Server** section. The gateway launches extensible-mcp as a `uvx` subprocess whose fastembed tool-search index uses an embedding model baked into the API image and loaded offline, so it no longer re-downloads that model from the Hugging Face Hub on every cold start. extensible-mcp is pinned twice over: to a commit, which fixes its own code, and with `--exclude-newer`, which fixes the resolution of its dependencies — without the cutoff uvx re-resolves them against PyPI on every networked start, so a new release of any of them would reach running deployments on restart. Both live in `_EXTENSIBLE_MCP_LAUNCH_ARGS` in `orchestrator/mcp_gateway.py`, the Dockerfile's pre-warm uses the same arguments (a unit test enforces the match), so an image whose pre-warm succeeded starts the gateway with no network. The pre-warm is best-effort (`|| true`): an image built while GitHub was unreachable fetches the pinned commit on its first start instead. Updating the gateway is a deliberate bump of the commit and cutoff; `uv` itself is not pinned. Because the MCP stdio client does not inherit the parent environment, `MCPGateway.start()` explicitly forwards the embedding-cache/offline env vars (`FASTEMBED_CACHE_PATH`, `HF_HUB_OFFLINE`, …) into the child.\n\n**Emailing an artifact.** A Gmail send or draft may carry `attachments=[{\"artifact_id\": \"alert:12\", \"as\"?: \"docx\"}]`. `call_tool` runs the recipient gate first; only if it passes does `_expand_artifact_attachments` replace each such entry with the rendered file (`content` base64 + `filename` + `mime_type`, workspace-mcp's own attachment shape), produced by `artifact_records.render_artifact_file` — the same renderer as `/artifacts/{id}/download`. Only artifact rows resolve (never an arbitrary alert), links have no file, and an entry mixing `artifact_id` with other keys is refused. A stringified or single-object `attachments` is normalised first, so an artifact entry can never reach workspace-mcp unexpanded (it would silently drop it). At most 5 distinct artifacts per email (repeats are dropped) and 15 MB rendered in total; rendering runs off the event loop via `asyncio.to_thread`. A refusal writes `artifact_attachment_refused`; a successful send writes `artifact_attached` — both carry the recipients. Other attachment entries pass through unchanged. The same expansion serves the Microsoft 365 mail writes: the entry becomes a Graph `#microsoft.graph.fileAttachment` (`name`, `contentType`, `contentBytes`) placed where Graph expects it for that tool (`body.Message.attachments` for `send-mail`, reply, forward and the createReply drafts; `body.attachments` for `create-draft-email`), whether the model put it at the top level or inside the message; an entry anywhere else, one on `update-mail-message` or `send-draft-message` (their requests cannot carry attachments), or one behind a body that is not an object is refused rather than forwarded unexpanded.\n\n**Enablement, and why it degrades instead of dying.** `MCP_ENABLED` left unset is inferred from the presence of `mcp_servers_config_path` — dropping an `mcp_servers.json` next to `profile.yaml` turns MCP on with no further configuration. An **explicit** setting always wins in both directions: `MCP_ENABLED=false` keeps MCP off with the file still in place. It did not: `_resolve_mcp` tested `not self.mcp_enabled`, which cannot tell \"never set\" from \"set to false\", so the file's presence silently overrode the operator. The file is probed through `config.mcp_config_file_present`, never `Path.exists` directly — that re-raises an EACCES on a parent directory, and from inside a validator that comes out of `Settings()` itself, which `api/main.py` calls at module level. The check now reads `model_fields_set` (the fields the env actually supplied) and stores that on a private attr behind the read-only `mcp_auto_enabled` property, because the validator's own assignment joins that set and would otherwise make every auto-enable look explicit downstream.\n\nStartup is **best-effort**, in `main._start_mcp_gateway`. Two checks come before the subprocess: MCP off is a no-op, and a config defining no servers under `mcpServers` is logged and skipped — extensible-mcp's own loader raises on that and exits, and because the child is already gone the only thing that surfaces is anyio's \"Attempted to exit cancel scope in a different task\" from `stdio_client` unwinding, a traceback naming nothing about configuration. The start is also bounded (120s, generous because that child may still clone extensible-mcp on a cold start if the image's pre-warm failed): `ClientSession` carries no `read_timeout_seconds`, so `initialize()` waits forever on a child that is *alive but silent*, and an unbounded wait is a lifespan that never yields — the same healthcheck-never-passes symptom, which no `except` can see. A gateway that fails anyway is logged at ERROR with the config path, the servers it named, and whether the config file or an explicit `MCP_ENABLED` turned MCP on, then torn down under its own timeout; the API boots **without MCP tools and without the email poller** rather than refusing to boot. The guard catches `BaseException`, not `Exception`, re-raising a bare `CancelledError` first: the original failure came from anyio, which wraps a task group's failures in `BaseExceptionGroup` as soon as one is a `BaseException` — and that group is not an `Exception`, so `except Exception` would miss the very shape this contains. The teardown is contained too, since it runs from inside that `except` and would otherwise replace the actionable log with the crash it prevents. `configured_server_names` never raises either, and deliberately catches broadly: it reads an operator-owned path during startup, where an exception is the crash loop rather than a bug report — `Path.is_file` re-raises an EACCES on a parent directory, and `json.loads` answers deeply-nested input with `RecursionError`, which is not a `ValueError`. It tests `is_file()` so a directory or a symlink to a FIFO is never opened, and caps the read at 1 MiB. That exception used to propagate out of the lifespan and kill the container — under `restart: unless-stopped`, a crash loop whose only outward symptom was a healthcheck that never passed. Outbound tools: per-Person DMs (`send_slack_dm`, `send_discord_dm`, `send_telegram_message`) plus room-targeted `send_department_message` and `send_company_broadcast`. All three per-Person DMs refuse a recipient id that matches no non-archived Person row (`schedule_tools._dm_recipient_on_roster`), which is what stops untrusted content in a turn — inbound mail, a chat message, a knowledge document — from directing a DM at an arbitrary workspace member. `send_slack_dm` previously consulted the roster only when the turn was already private to the principal, so on an ordinary turn it applied no roster check at all while its Discord and Telegram siblings did; the room-targeted tools are deliberately ungated, since those channels are team-visible. The per-Person DMs are additionally rate-limited and de-duplicated by the outbound anti-spam guard (below).\n\n### Outbound anti-spam guard\n\nEvery per-Person DM passes a deterministic guard (`orchestrator.outbound_guard.check_outbound_allowed`) after the roster gate and recipient-id recovery but before the network send — the single chokepoint **all** proactive DM paths reach (ad-hoc chat sends, the scheduler's followup/nudge dispatch, principal-brief delivery, the reflection turn, and research-finding routing). On a violation it returns a human-readable reason and the handler **suppresses** the send (`{\"status\":\"suppressed\",\"reason\":…}`), writes a `tool_invocation` audit row with `suppressed=true`, and writes **no** `done` activity row — so a blocked attempt never counts toward its own rate cap, and the Executive can read the reason and hold or `schedule_followup` for the recipient's next window. Three controls: **content dedup** (a near-identical message — whitespace/case-normalized, first 160 chars — already sent to the same recipient within `OUTBOUND_DEDUP_WINDOW_MINUTES`, default 360); a **per-recipient rate cap** (`OUTBOUND_MAX_PER_RECIPIENT_PER_WINDOW`, default 5, over `OUTBOUND_RATE_WINDOW_MINUTES`, default 60); and **quiet hours / availability** (`OUTBOUND_RESPECT_QUIET_HOURS`, default on — refuse while the recipient is on leave or outside their `people.channel` availability windows, surfacing the next open window). Delivery history is read from the existing `status='done'` `scheduled_actions` rows (`episodic.recent_sends_for_channel_ref`), so there is no new table. The guard is **fail-open**: any internal lookup error returns \"allow\", so it can never turn a working send into a crash or a silently-dropped message. Principal briefs and the internal heartbeat kinds use the `__internal__` channel and never reach these handlers, so they are unaffected.\n\n### AI disclosure\n\nWhat the Executive sends says it is an AI assistant (`orchestrator.disclosure`), added by plain code so a reply the model writes can't leave it out. **Mail** from its own mailbox gets a line under the body, by default \"<EXEC_DISPLAY_NAME> is an AI assistant working for <company>. To reach a person, reply and ask, or write to <principal's address>.\": the MCP gateway adds it to every Gmail send and draft and every Outlook send, reply, forward and draft (`mcp_gateway._with_disclosure`) **after** every outbound gate has passed and only on what goes to the server, so the gates, the outbound-context record and the audit see the model's own words, and `send_from_executive` gets it the same way. It counts as already there only when it is the last line that shows (in a quoted earlier message or hidden in markup it doesn't), and a send of an existing draft and mail whose every recipient is the principal go as they are. HTML that could hide what follows (an unclosed comment, a style or script block, text styled out of sight: `disclosure.hides_what_follows`) is refused, so the line can't be hidden after it; the model is told to send plain text. The roster acknowledgement to a new sender, the one send to an address nobody checked, gets `BARE_MAIL_LINE` (\"This message was sent by an AI assistant.\"), which names no one, like the acknowledgement itself. **Chat** gets a shorter line (\"<name> · AI assistant\", in italics) on the Executive's replies in Slack, Telegram, Discord and Google Chat and on its own DMs and room posts (`send_slack_dm`, `send_discord_dm`, `send_telegram_message`, `broadcast_tools._dispatch`), but only on the first message in a conversation and again after 24 hours without one (`with_chat_line`, kept in memory, so a restart shows it once more); session history keeps the reply without it. `AI_DISCLOSURE` (default on) turns it off; `AI_DISCLOSURE_MAIL_TEXT` / `AI_DISCLOSURE_CHAT_TEXT` reword the lines, and a wording that doesn't say \"AI\" is ignored for the built-in one. It fails open: a line that can't be built is left off, never a send refused. Replies sent *as a person* under Act as me are the delegation section's (**AI line on replies**).\n\n### Calendar & video links\n\nThe Executive books meetings through two typed tools over the configured calendar backend (`CALENDAR_PROVIDER`: Google Calendar via `manage_event`, or Outlook via `create-calendar-event` — see *Workspace providers*). `create_calendar_event` schedules a **future** meeting (propose→approve by default, or auto-execute once that decision class is promoted) and may be used proactively when a stalled goal warrants a sync. `create_instant_meeting` spins up an **impromptu call starting now**, auto-executing without the approval step and skipping the business-hours/horizon caps, so the Executive can hand back a join link immediately. Because nothing approves it, an instant meeting, and `cancel_calendar_event` too, runs only when someone on the team asked on a surface that verified who they are (the principal, or a rostered teammate in the app or their own chat); a turn an inbound email started gets a refusal to pass on instead, and a team-mode `create_calendar_event` the operations department would auto-execute is proposed instead. Both attach a video-meeting link by default (`add_video_link`; opt out for in-person) — a **Google Meet** link parsed from the created event's `conferenceData`, or a **Microsoft Teams** link from `onlineMeeting.joinUrl` — surfaced to the caller plus persisted on the decision ledger as `meet_link`. The tool descriptions are static, backend-neutral text, so the cached tool block never depends on settings; the pre-rename `add_google_meet` key is still honoured for old ledger rows. Either way the caps that matter still hold — roster membership, principal-protection, max attendees, and a daily booking ceiling.\n\n**Around-the-meeting participation.** The Executive doesn't join the call as a live voice; instead, on every successful booking it schedules a best-effort post-meeting follow-up (~5 min after the end) that DMs a human attendee asking for a quick recap — key decisions and action items. The reply flows back through the same cross-session DM context linkage described below, so the Executive can act on it (summarize, follow up) in a later turn. No meeting-transcript or Drive integration is required.\n\n### Workspace providers\n\nThe Executive can *discover* any mail or calendar tool the gateway exposes, but three code paths call a fixed backend with no model in the loop and used to spell Gmail tool names inline: the inbound mailbox poller (list unread → fetch → mark read), alert email dispatch plus the scheduler's email send-tool hint, and the typed booking tools plus the approval-time conflict check. `integrations.workspace` puts those behind two Protocols (`MailProvider`, `CalendarProvider`) with one module per backend (`google.py`, `microsoft.py`) and a registry keyed on `EMAIL_PROVIDER` / `CALENDAR_PROVIDER` — separate switches (Outlook mail with a Google calendar is legitimate), both defaulting to `google`, so an existing install changes nothing until an operator opts in. The Google backend is the old inline code moved verbatim; tests pin its argument dicts byte-for-byte. The Microsoft backend polls only the Inbox folder (`list-mail-folder-messages` with `isRead eq false`, never the all-folders listing that includes Junk), converts Graph's HTML bodies to text with a real parser, lists a message's attachments so the names appear in the same `N. name (mime, size KB)` shape Gmail's text uses (and reach peer memory), marks read with `update-mail-message`, sends through `send-mail`'s nested shape, books with `isOnlineMeeting` + `onlineMeetingProvider=teamsForBusiness`, cancels with `delete-calendar-event`, and answers the advisory conflict check from the Executive's **own** calendar view (delegated permissions cannot read attendees' free/busy — narrower than Google's `query_freebusy`).\n\nEvery backend renders an inbound message into the same `Subject:` / `From:` / `To:` / `Cc:` + `--- BODY ---` text the persona already expects, and the poller adds a `--- REPLY ---` block naming the exact reply tool and threading ids for *that* message (Gmail: `send_gmail_message` + `thread_id`; Outlook: `reply-mail-message` + `messageId`) after the mail's `<untrusted_content>` block, never inside it. The block rides in the user turn — which is what lets the cached persona say \"use the reply tool in the REPLY block\" instead of naming a vendor tool — and the scheduler's proactive-email trigger names the backend's send tool the same way. Providers never raise on a malformed backend response (they return empty / `None` / `{\"error\": …}`), so one bad cycle never kills a loop, and a chosen backend whose MCP server is missing from `mcp_servers.json` is logged at startup and skipped by the poller: fail-soft, never a boot failure. The poller's newer checks run on that rendered text for both backends (contacts and private turns, roster requests and their emailed answers, the audit row), with three Outlook differences that fail closed: no sender is ever authenticated (the `dmarc=pass` read is Gmail's), so there is no acknowledgement to an unknown sender, no emailed fact confirmation and no principal-authenticated mail; attachments stay a list (the download is Gmail's); and the brief's calendar block is skipped (its listing is Google Calendar's). The code's own emails to the principal (roster prompts, fact confirmations, scheduled deliveries) go through `workspace.registry.send_from_executive`, and `runner.email_ready` follows `EMAIL_PROVIDER`. A private turn's allow-list (`schedule_tools.PRIVATE_TURN_MCP_TOOLS`) and Act as me's mail-touched reads include the Outlook twins of the Gmail and Calendar tools, and roster answer tokens are hidden from Microsoft 365 results as from Google's.\n\n### Cross-session DM context\n\nWhen the principal says \"DM Alex about X\" in one session, OE sends the DM and — if a live conversation is active and the recipient is not one of the principal's contacts (a contact cannot DM the bot, and quoting the principal's conversation into their email reply's unattended turn would put it in the audit log) — records an `outbound_context` linkage row (recipient, full outbound text, originating session id). Email is covered too but recorded one layer deeper — there is no `send_email` tool, so the linkage is written in `orchestrator.mcp_gateway` after a successful `google_workspace__send_gmail_message` or `microsoft_365__send-mail`, one row per `to`/`cc` recipient keyed by the bare lowercased address (OE's own address skipped). When Alex later replies on the 1:1 DM path — or by email — the shared helper `integrations.inbound_hydration.hydrate_user_message` (called by the Discord, Telegram, and Slack handlers, and by the email poller keyed on the sender address) looks up the most-recent `open` linkage for that sender within 72h and prepends an `<outbound_reply_context>` block — the outbound text plus a backstory excerpt from the originating session — to the message handed to `Executive.chat()`. So OE knows what the reply is about instead of starting from an empty per-person session. The block is context for the LLM turn only: peer memory strips it before recording what the person said (`strip_outbound_reply_context`, applied in `sync_turn`, `sync_department_turn` and to the prefetch query), so the Executive's own DM is never attributed to the person who replied to it. The block lands in the user turn (caching rule) and is injected into a copy only — the pristine reply is persisted to history. The linkage is one-shot (consumed on first matched reply, race-safe) so an unrelated later DM isn't re-hydrated. This is deterministic and independent of the per-person `peer_memory` (Honcho) layer.\n\nEvery adapter that persists a turn (Slack, Discord, Telegram, and the web route) records the message's **resolved sender** on `chat_messages.sender_person_id` — the rostered Person the adapter matched, or NULL for an unrostered sender — independently of the session owner, which falls back to the principal on Discord and Telegram so those threads stay visible. Attunement's open loops and feedback rely on this so an outsider's words are never read as someone on the roster's.\n\n### OneDrive folder sync\n\nThe Microsoft twin of the Drive folder sync below (`knowledge.onedrive_sync`, `ONEDRIVE_SYNC_ENABLED`; setup in `docs/onedrive_sync_setup.md`). It copies the files in `ONEDRIVE_SYNC_FOLDERS` (`<drive id>/<item id>` pairs; `openexecutive onedrive-folder <link>` resolves a sharing link) into the isolated `onedrive_docs` collection on its own heartbeat, and shows them on the knowledge page as \"OneDrive\" rows with a **Sync now** button. Microsoft has no service account that sees only what is shared with it (app-only Graph access covers the whole tenant), so it reads as the Executive's own Microsoft 365 sign-in, and the folder list is the scope. It never stores a credential: each tick asks the launcher for a token (`ms365-mcp-launch.sh --access-token`, which runs `docker/ms365-access-token.mjs` through the server's own token cache and will only mint file-read tokens), keeps it in memory, and sends it only to Graph. `knowledge.onedrive_client` issues GETs only, follows a paging link only back to Graph, and fetches a file's pre-authenticated download link without the token; errors never carry that link. PowerPoint and older Office formats are read through Graph's PDF rendering.\n\n### Google Drive folder sync\n\nThe synced files show on the knowledge page's Company documents list as \"Google Drive\" rows (sync time, last edit in Drive, an Open in Drive link and a text view), and the page's source card shows the last run, any sign-in error or count of files that failed to sync, and a **Sync now** button that runs one tick on demand.\n\n\nSeparate from the Drive tools the gateway proxies. The optional folder sync (`knowledge.drive_sync`, `DRIVE_SYNC_ENABLED`; setup in `docs/drive_sync_setup.md`) copies the files in `DRIVE_SYNC_FOLDER_IDS` into the isolated `drive_docs` knowledge collection on a scheduler heartbeat, so the Executive knows those folders without searching Drive live (see RAG). It calls the Drive v3 API directly (`knowledge.drive_client`, `httpx`) as a **service account** with the `drive.readonly` scope (`DRIVE_SYNC_SERVICE_ACCOUNT_FILE`), not as the Executive's own account through workspace-mcp. Google enforces the folder scope, because the account sees only what is shared with it, and the sync can change nothing. Structured JSON means no display text to parse, and every file or folder id is checked against `[A-Za-z0-9_-]{1,128}` before it goes into a URL or a query. The two paths meet in two places. A synced chunk carries its Drive file id and sync time, so the Executive can open the live file with its own `get_drive_file_content`. The setup doc says to share each synced folder with the Executive's account too, so that live open works.\n\n### Confluence space sync\n\nThe optional Confluence sync (`knowledge.confluence_sync`, `CONFLUENCE_SYNC_ENABLED`; setup in `docs/confluence_sync_setup.md`) copies the pages of the spaces in `CONFLUENCE_SYNC_SPACE_KEYS` into the isolated `confluence_wiki` knowledge collection on a scheduler heartbeat, so the Executive knows the company wiki without a tool call (see RAG). It works with Confluence Cloud and Server/Data Center: `knowledge.confluence_client` calls the v1 REST API both serve, directly with `httpx`, signing in with a personal access token (Server/DC) or a username and API token (Cloud). It reads only, never follows a redirect (so the token stays on the configured site), streams responses with a 20 MB cap, and checks page ids and space keys before they reach a URL or a CQL query. The URL must be `https://` unless `CONFLUENCE_SYNC_ALLOW_HTTP` is set, and `CONFLUENCE_SSL_VERIFY` takes `true`, `false` or a CA bundle path for a private certificate authority. With `CONFLUENCE_SYNC_PUBLIC_HOSTS_ONLY` (off by default, for an install where whoever sets the URL must not reach the backend's own network) every request resolves the site's host, refuses it when any address is not public (loopback, private, link-local, shared), and connects to the checked address with the site's name kept for the Host header and the TLS certificate check, directly rather than through a proxy. `confluence_client.config_problem` is the same settings check the app runs at start, for a caller that writes these settings on someone's behalf.\n\nIt is separate from a Confluence MCP server such as mcp-atlassian, which a deployment can add to `mcp_servers.json` for live reads and edits in chat. The two share the env names `CONFLUENCE_URL`, `CONFLUENCE_PERSONAL_TOKEN`, `CONFLUENCE_USERNAME`, `CONFLUENCE_API_TOKEN` and `CONFLUENCE_SSL_VERIFY`, so one set of values serves both, and a synced chunk carries its page id so the Executive can open the live page with such a tool. A Confluence token is a user's, and what the sync reads becomes retrievable by everyone who talks to the Executive, so the space list is required and pages with a read restriction (their own or a parent's) are skipped unless `CONFLUENCE_SYNC_SKIP_RESTRICTED=false`. That keeps any existing member's token safe to use, without a dedicated licence. Manual runs: `openexecutive sync-confluence` / `purge-confluence`.\n\n### Your own Gmail or Outlook (Act as me)\n\nA second, separate mail connection, for **Act as me** (see **Act as Me (Delegation)**): a person's *own* mailbox, which `delegation.gmail.DelegateGmail` reads and drafts into over the Gmail REST API. It is never registered with the MCP gateway, so the model can reach it only through the typed `ghostwrite_email` handler — never by tool name through `call_tool`. The credential is one file per person in `DELEGATION_GOOGLE_CREDENTIALS_DIR` (default `company/delegation_google/`, `/data/delegation_google` in Docker), minted on a computer with a browser by `scripts/connect-own-gmail.py --email <you>` with the Executive's own OAuth client and the scopes `gmail.readonly` and `gmail.compose` — the same seed-it-yourself flow as the Executive's token, since the API serves no OAuth callback. It never lives in `WORKSPACE_MCP_CREDENTIALS_DIR`, where workspace-mcp picks a credential by address.\n\nEvery use re-checks with Google that the credential opens the mailbox on the person's People entry, and refuses outright when that address is the Executive's own: a shared mailbox would make everything the Executive writes look like theirs. Drafts are threaded (`In-Reply-To`, `References` of whole ids ending with the parent, `threadId`) and stamped `X-OE-Ghostwritten`; the only send is `drafts.send` of an existing draft by id, when the person taps Send on its reply card (`delegation.reply_send`). A Gmail rate limit, even as a 403, is an error to retry rather than a reconnect. The voice learner and the thread shown to the ghostwriter keep only each sender's own words through `email_poller.sender_new_text`, the poller's quoted-reply stripper made public. Setup status shows it as **Your own mailbox (Act as me)**.\n\nThe same file can instead hold an Outlook sign-in (`\"provider\": \"microsoft\"`), minted by `scripts/connect-own-outlook.py --email <you>` with a device-code sign-in on the Executive's own Entra app (`MS365_MCP_CLIENT_ID`, `MS365_MCP_TENANT_ID`, default `common`) and the delegated scopes `User.Read`, `Mail.ReadWrite`, `Mail.Send` and `offline_access`. `delegation.gmail.gmail_for` then hands back `delegation.outlook.DelegateOutlook`, which talks to Microsoft Graph directly — never the Microsoft 365 MCP server the Executive's own Outlook uses — with the same methods and the same one send: Graph's `send` on an existing draft, with no body. Microsoft rotates the refresh token on use and the new one is written back into the file. A sender counts as verified only from the topmost `Authentication-Results`, the one Exchange wrote, and only mail in their Sent Items counts as theirs.",
"mermaid": "flowchart TD\n Channels[\"Slack / Discord<br/>Telegram / GChat / Email\"]\n Resolve[\"find_person_by_*_id\"]\n Gate[\"roster gate\"]\n Drop[\"drop\"]\n Hold[\"roster request<br/>(hold + fixed ack,<br/>ask the principal)\"]\n Exec[\"Executive.chat()\"]\n Attach[\"attachments<br/>(vision / inbound_attachments)\"]\n MCP[\"MCP gateway\"]\n GW[\"Google Workspace<br/>(workspace-mcp child)\"]\n M365[\"Microsoft 365<br/>(ms-365-mcp-server child:<br/>Outlook + OneDrive)\"]\n WP[\"workspace providers<br/>(poller / alerts / booking)\"]\n Guard[\"outbound anti-spam guard<br/>(dedup / rate cap / quiet hours)\"]\n Out[\"DM / department<br/>/ broadcast tools\"]\n Channels --> Resolve\n Resolve --> Gate\n Gate -- \"unknown sender\" --> Hold\n Hold -- \"approved: replay\" --> Exec\n Hold -- \"contact / declined\" --> Drop\n Gate -- \"Person\" --> Exec\n Gate --> Attach\n Attach --> Exec\n Exec --> MCP\n MCP -- \"roster egress gate\" --> GW\n MCP -- \"roster egress gate\" --> M365\n WP -- \"EMAIL_PROVIDER / CALENDAR_PROVIDER\" --> MCP\n WP --> Exec\n Exec --> Guard\n Guard -- \"suppress\" --> Drop\n Guard -- \"allow\" --> Out\n Out --> Channels\n Own[(\"Principal's own Gmail<br/>(Act as me: REST, drafts only)\")]\n Exec -- \"ghostwrite_email\" --> Own\n class Channels entry\n class Resolve compute\n class Gate compute\n class Drop external\n class Hold compute\n class Exec hot\n class Attach compute\n class MCP external\n class GW external\n class M365 external\n class WP compute\n class Guard compute\n class Out compute\n class Own external",
- "generated_at": "2026-10-12T19:17:47Z"
+ "generated_at": "2026-10-13T15:30:00Z"
}
diff --git a/packages/core/openexecutive/architecture/prebuilt/schemas.json b/packages/core/openexecutive/architecture/prebuilt/schemas.json
index 7b7bbee3..f6cbc1d7 100644
--- a/packages/core/openexecutive/architecture/prebuilt/schemas.json
+++ b/packages/core/openexecutive/architecture/prebuilt/schemas.json
@@ -1,7 +1,7 @@
{
"section_id": "schemas",
"title": "Data Schemas",
- "markdown": "The shapes below are the load-bearing data structures: episodic SQLite tables (in `episodic_memory.db`) and the Pydantic models for departments and people.\n\n### Episodic tables\n\n| Table | Key columns |\n|---|---|\n| `decisions` | `id`, `timestamp`, `domain`, `summary`, `rationale`, `outcome`, `tags`, `department`, `session_id` |\n| `initiatives` | `id`, `title`, `status`, `created_at`, `updated_at`, `summary`, `department` |\n| `advice_given` | `id`, `timestamp`, `domain`, `query_summary`, `advice_summary`, `department`, `session_id` |\n| `facts` | `id`, `kind` (fact / correction / profile), `subject`, `subject_key`, `statement`, `previous_statement`, `source_quote`, `source_channel`, `session_id`, `turn_id`, `recorded_by_person_id`, `created_at`, `status` (active / superseded / retired), `superseded_by`, `retired_at`, `retired_reason` |\n| `scheduled_actions` | `id`, `created_at`, `run_at`, `channel`, `channel_ref`, `intent_text`, `status`, `attempts`, `kind`, `department`, `assigned_to_person_id`, `awaiting_response_since`, `scope_key`, `required_scope` |\n| `outbound_context` | `id`, `created_at`, `channel`, `channel_ref`, `recipient_person_id`, `outbound_text`, `originating_session_id`, `outbound_message_id`, `consumed_at`, `status` |\n| `history_reminders` | (`person_id`, `day`) PK, `sent_at` — the days a person was reminded of notes due (`memory.history_reminders`), older days pruned on write |\n| `session_drive_files` | (`session_id`, `person_id`, `file_id`) PK, `name`, `mime_type`, `link`, `summary` (first 600 chars of an opened file), `opened`, `found_by` (the search query), `first_seen_at`, `last_seen_at` — Drive files a conversation found or opened (`memory.drive_reads`) |\n| `session_drive_searches` | `id`, `session_id`, `person_id`, `query`, `result_count`, `searched_at` — every Drive search a conversation ran |\n| `saved_tools` | `name` (PK), `description`, `enabled`, `current_version`, `workflow_version` (the version workflows may run, NULL until the owner turns it on; never set for a Python tool), `kind` (`script` for `run_script`, `python` for a kept `run_python_job`; added with default `script`), `created_at`, `updated_at` — scripts and Python jobs the Executive saved to run again (`workflows.saved_tools`) |\n| `saved_tool_versions` | (`name`, `version`) PK, `description`, `script`, `tools` (JSON: the tools the script called when saved), `origin` (`chat` or `workflow:<name>/<step>`), `created_at` |\n| `saved_tool_runs` | `id`, `name`, `version`, `ok`, `calls`, `duration_ms`, `origin`, `at` — the newest 200 runs per tool |\n| `chat_messages` | `id`, `session_id`, `role`, `content`, `created_at`, `action_chips`, `stopped`, `sender_person_id`, `feedback`, `feedback_note`, `feedback_by_person_id`, `sources` (JSON `{sources: [{kind, title, url}], unavailable: [area]}` — what an assistant reply looked at and which areas it left out; NULL on older rows) |\n| `attunement_usage` | `day` (PK, UTC date), `calls` — the per-day ceiling on open-loop extraction model calls |\n| `attunement_profiles` | `person_id` (PK), `rules` (JSON: up to 4 `{text, basis, evidence}`), `locked`, `updated_at`, `updated_by`, `last_pass_at`, `pass_day`, `passes_today`, `last_message_id` — per-person working style plus pass pacing |\n| `attunement_profile_history` | `id`, `person_id`, `created_at`, `rules`, `locked`, `updated_by` — every change to a working style |\n| `proactive_outcomes` | `id`, `created_at`, `person_id`, `source`, `ref`, `channel`, `channel_ref`, `outbound_context_id`, `outcome` (`replied` / `acted` / `ignored` / `void`, NULL while open), `resolved_at` — one row per proactive DM (Attunement outcome ledger) |\n| `delegation_settings` | `person_id` (PK), `enabled` (default 0), `updated_at`, `updated_by` — Act as me, per person; no row means off |\n| `history_notes` | `id`, `person_id`, `visibility` (`private`), `source` (`approved_reply`), `channel` (`email`), `conversation_key` (a hash of the channel and thread, never the thread id), `counterpart`, `subject`, `kind` (`promised` / `agreed` / `declined` / `answered` / `asked` / `shared`), `summary`, `quote` (the person's own words), `due_date`, `trust`, `occurred_at`, `created_at`, `expires_at` (NULL: until forgotten), `pinned`, `correction`, `corrected_at`, `kept_from` (when the retention clock started: NULL means `occurred_at`; an unpin restarts it) — Always in the loop, one row per note |\n| `history_settings` | `id` (1), `retention_days` (NULL: until forgotten; no row means 90), `updated_at`, `updated_by` — the company retention |\n| `history_person_settings` | `person_id` (PK), `reply_notes` (default 0), `retention_days` (NULL: the company's), `updated_at`, `updated_by` |\n| `history_excluded` | `person_id`, `conversation_key` (PK together), `created_at` — conversations a person asked not to remember |\n| `history_chat_passes` | `person_id`, `day` (PK together, UTC `YYYY-MM-DD`), `passes` — note passes run on a person's chat messages that day, against the daily cap; older days pruned on write |\n| `delegation_voice` | `person_id` (PK), `profile` (JSON: `greetings` per audience, `sign_off`, `signature`, `length`, `formality`, up to 8 `habits`, up to 6 `avoid`, up to 3 `exemplars`), `locked`, `learned_at`, `sample_count`, `updated_at`, `updated_by` — \"How I write\" |\n| `delegation_voice_history` | `id`, `person_id`, `created_at`, `profile`, `locked`, `updated_by` — every change to a writing profile |\n| `delegation_drafts` | `id`, `person_id`, `source` (`chat` \\| `inbox`), `thread_id`, `draft_id`, `message_id`, `sent_message_id`, `created_at` — one row of ids per draft Act as me saved in someone's Gmail (no address, subject or text): the daily limit counts it, and \"How I write\" skips a chat draft's thread and a sent draft's message |\n| `delegation_inbox_watch` | `person_id` (PK), `enabled`, `watch_since` (reset on every off→on), `last_poll_at`, `status`, `backoff_until`, `failures`, `mailbox_drafts` (also save each reply in their Drafts; default 0, the reply lives on its card alone), `updated_at`, `updated_by` — **Draft replies to my inbox**, per person; no row means off |\n| `delegation_inbox_messages` | `id`, `person_id`, `message_id`, `thread_id`, `received_at`, `sender_key` (a hash of the address, never the address), `relation`, `outcome` (`processing`, `retry` — left for a later scan, `skipped`, `not_needed`, `deferred`, `failed`, `drafted`, `sent`, `dismissed`, `closed`, `expired`), `reason`, `classified`, `attempts`, `draft_id`, `decision_id`, `flags` (JSON), `created_at`, `updated_at`; `UNIQUE(person_id, message_id)` — the inbox watcher's cursor: what happened to each message it looked at |\n| `delegation_handle_it` | `person_id` (PK), `enabled`, `mode` (`careful`, `balanced` or `bold`), `levels` (the per-kind setting `mode` replaced, kept for older rows), `updated_at`, `updated_by` — **Handle it for me**, per person; no row means off. A `training` mode from before per-setting training moves to `balanced` with Replies and Follow-ups in training |\n| `delegation_training` | `person_id`, `setting` (`replies`, `follow_ups`, `actions` or `drafts`), `enabled`, `updated_at`, `updated_by`; primary key `(person_id, setting)` — Act as me settings in training, per person; no row means not in training |\n| `delegation_handled` | `id`, `person_id`, `thread_id`, `decision_id`, `sent_message_id`, `sent_at`; index on `(person_id, sent_at)` — one row per reply or follow-up Handle it for me sent on its own, read for its one-per-thread-a-day and daily limits |\n\n| `dynamic_workflows` | `name` (PK), `definition` (JSON of `DynamicWorkflowDef`), `is_active`, `created_at`, `updated_at`, `owner_person_id` (creator; set on insert only, kept out of the definition JSON) |\n| `workflow_approved_targets` | (`workflow_name`, `value`) PK, `key`, `approved_at`, `run_id` — targets a workflow's tool steps may write to without asking again; cleared when the workflow is deleted |\n| `executive_control` | `id` (PK, always 1), `paused`, `paused_at`, `paused_by`, `reason`, `updated_at` — the operator pause switch (`scheduler.pause`); operator-level, so client-slot switches carry it across |\n| `app_migrations` | `name` (PK), `applied_at` — one-shot data migrations (sweeps that must run exactly once per DB); schema DDL stays idempotent and is not recorded here |\n\n`department` and `session_id` are additive columns (default `''`); `session_id` scopes per-thread recall. `scheduled_actions.status` ∈ pending / running / done / failed / cancelled, and `scheduled_actions.kind` includes `dynamic_workflow` (a cadence-fired user-created workflow run) and `open_loop` (an Attunement open loop: `assigned_to_person_id` is the owner, `awaiting_response_since` the due time — cleared when the loop closes — and `scope_key` `loop:{owner}:{hash}`, unique among open loops via the partial index `idx_scheduled_open_loop_scope`). `chat_messages.sender_person_id` is the resolved rostered sender of a user message (never the session owner's principal fallback; NULL on legacy rows and for unrostered senders); `feedback` is `up` / `down` on an assistant reply. `dynamic_workflows` stores user-created (dynamic) workflow definitions interpreted by the generic `DynamicWorkflow` engine; the JSON `definition` holds free-text input fields and an ordered step list (specialist / approval_gate / synthesis) plus an optional cadence.\n\n### Legacy tables (talent / staff onboarding)\n\nThe talent (executive-search) and staff-onboarding features were removed. Their tables —\n`engagements`, `candidates`, `offers`, `onboarding_templates`, `onboarding_plans`,\n`onboarding_tasks` — are **not dropped**: no code creates, reads or writes them any more,\nbut rows survive on upgraded installs. `clients.slots` still wipes them when a blank client\nslot is activated (existence-guarded, a no-op on fresh installs) so a slot can never inherit\na previous client's candidate data.\n\n### Department / Goal / ChecklistItem\n\n| Model | Fields |\n|---|---|\n| `DepartmentConfig` | `slug`, `title`, `specialist_key`, `charter`, `authority_level`, `head_person_id`, `head_persona_slug`, `cadences`, `slack_channel_id`, `discord_channel_id`, `telegram_chat_id` |\n| `Goal` | `id`, `department_slug`, `period_type`, `period_value`, `key_result`, `target`, `current`, `status`, `created_at`, `updated_at`, `last_reviewed_at` |\n| `DepartmentState` | `config`, `goals`, `headcount`, `budget_usd`, `member_person_ids`, `updated_at` |\n\n`authority_level` ∈ `auto_execute` / `propose_only` / `escalate`; `Goal.status` ∈ `on_track` / `at_risk` / `off_track`.\n\n### Person\n\n| Field group | Fields |\n|---|---|\n| Identity | `id`, `full_name`, `role`, `is_principal`, `kind`, `department_slugs` |\n| Contact | `email`, `email_aliases`, `slack_user_id`, `telegram_chat_id`, `discord_user_id`, `preferred_channel` |\n| Routing | `availability`, `authority_scope`, `response_sla_hours`, `on_leave_until`, `reports_to_person_id` |\n| Lifecycle | `archived`, `created_at`, `updated_at` |\n\nExactly one `Person` should have `is_principal=True` — the fallback approver when no delegated person matches.\n\n`kind` is `team` (the default) or `contact` — someone outside the team, such as a client or contractor. It is an additive column, `kind TEXT NOT NULL DEFAULT 'team'`, added by the same `PRAGMA table_info`-guarded migration as `discord_user_id`, so every existing row is a team member. The principal is always `team`: the store refuses to write a principal contact (`PrincipalContactError`), and `initialize_db` repairs a hand-edited one on boot. A stored value the code does not know reads as `contact`. Every roster read in `people.store` and `people.registry` returns team members only unless it passes `include_contacts=True` (see Org).\n\n`email` is the one address a person signs in with; `email_aliases` (read from the `person_emails` table, up to 10) are other addresses they write from — they match inbound mail and may be emailed, but never sign in (see Org).\n\n### People-side tables\n\nThese live next to `people` (created by `people.store.initialize_db`, which also runs `people.roster_requests.initialize_tables`):\n\n| Table | Key columns |\n|---|---|\n| `person_emails` | `id`, `person_id` (FK → `people`, `ON DELETE CASCADE`), `email`, `email_norm` (trimmed + lowercased, **UNIQUE** across the roster), `source` (`manual` / `chat_tool` / `roster_request`), `roster_request_id`, `created_at` — a person's aliases; deleted when they are archived |\n| `roster_requests` | `id`, `channel` (`email` / `slack` / `discord` / `telegram`), `channel_ref` (the address or chat id), `display_name` (sanitised, unverified), `profile_email`, `on_company_domain`, `suggested_kind`, `suggested_person_id`, `status` (`pending` / `resolving` / `approved` / `linked` / `declined` / `expired` / `superseded`), `resolved_person_id`, `resolved_kind`, `resolved_via` (`web` / `slack` / `discord` / `telegram` / `email` / `people_page`), `resolved_at`, `token_hash` (the email answer token, hashed; cleared on resolve), `confirm_message_id`, `notified_at`, `notified_channel`, `ack_sent_at`, `alert_id`, `message_count`, `first_seen_at`, `last_seen_at`, `expires_at` — one pending row per `(channel, channel_ref)` (partial unique index) |\n| `roster_request_messages` | `id`, `request_id` (FK, cascade), `external_id` (unique per request, so a redelivery is held once), `payload_json` (what the adapter needs to replay it), `preview`, `received_at`, `replay_status` (`held` / `replaying` / `replayed` / `dropped` / `failed` / `unavailable`), `replayed_at` — at most 10 per request; `payload_json` and `preview` are blanked once a message is replayed or dropped |\n| `roster_ack_log` | `id`, `channel`, `channel_ref`, `sent_at` — one row per acknowledgement sent to a new sender, for the per-sender window and the daily cap |\n\n**Take the lead tables** (`orchestrator.take_the_lead`, created on first use, per company in `clients.slots` and `reset_all_state`): `take_the_lead` (`scope` PK: `executive` or `person:<id>`, `enabled`, `training` (added in place on older databases), `updated_at`, `updated_by`; each kind's Ask first is a `decision_class_state` row for `take_the_lead_action:<kind>`); `take_the_lead_rules` (`id`, `scope`: `company` or `person:<id>`, `kind` person|domain|words|amount, `value`, `created_by`, `created_at`); `take_the_lead_log` (`id`, `at`, `scope`, `source` reflection|research|scheduled, `tool`, `summary`, `status` done|waiting|approved|declined|failed, `why`, `decision_id`). A held action is a `decision_instances` row of class `take_the_lead_action` whose payload carries the tool, its input, whether it is an MCP tool, the gate kind or rule, the reason and the summary. `take_the_lead_allowed` (`id`, `key` unique: `<feature>|<action>|<who>`, `label`, `feature`, `example` (JSON of the edited fields, or empty), `uses`, `last_used_at`, `created_by`, `created_at`, `decision_id`, `removed_at`: set by Remove, so only approvals after it count toward a new suggestion, `person_id`: whose it is for a feature that learns per person, Act as me's `act_as_me|person:<id>|<reply|follow_up|message|invite|style>|<who>`, NULL for Take the lead; added in place on older databases): what it may do on its own in training, at most `ALLOWED_MAX` (200) per feature and person.",
+ "markdown": "The shapes below are the load-bearing data structures: episodic SQLite tables (in `episodic_memory.db`) and the Pydantic models for departments and people.\n\n### Episodic tables\n\n| Table | Key columns |\n|---|---|\n| `decisions` | `id`, `timestamp`, `domain`, `summary`, `rationale`, `outcome`, `tags`, `department`, `session_id` |\n| `initiatives` | `id`, `title`, `status`, `created_at`, `updated_at`, `summary`, `department` |\n| `advice_given` | `id`, `timestamp`, `domain`, `query_summary`, `advice_summary`, `department`, `session_id` |\n| `facts` | `id`, `kind` (fact / correction / profile), `subject`, `subject_key`, `statement`, `previous_statement`, `source_quote`, `source_channel`, `session_id`, `turn_id`, `recorded_by_person_id`, `created_at`, `status` (active / superseded / retired), `superseded_by`, `retired_at`, `retired_reason` |\n| `scheduled_actions` | `id`, `created_at`, `run_at`, `channel`, `channel_ref`, `intent_text`, `status`, `attempts`, `kind`, `department`, `assigned_to_person_id`, `awaiting_response_since`, `scope_key`, `required_scope` |\n| `outbound_context` | `id`, `created_at`, `channel`, `channel_ref`, `recipient_person_id`, `outbound_text`, `originating_session_id`, `outbound_message_id`, `consumed_at`, `status` |\n| `history_reminders` | (`person_id`, `day`) PK, `sent_at` — the days a person was reminded of notes due (`memory.history_reminders`), older days pruned on write |\n| `session_drive_files` | (`session_id`, `person_id`, `file_id`) PK, `name`, `mime_type`, `link`, `summary` (first 600 chars of an opened file), `opened`, `found_by` (the search query), `first_seen_at`, `last_seen_at` — Drive files a conversation found or opened (`memory.drive_reads`) |\n| `session_drive_searches` | `id`, `session_id`, `person_id`, `query`, `result_count`, `searched_at` — every Drive search a conversation ran |\n| `saved_tools` | `name` (PK), `description`, `enabled`, `current_version`, `workflow_version` (the version workflows may run, NULL until the owner turns it on; never set for a Python tool), `kind` (`script` for `run_script`, `python` for a kept `run_python_job`; added with default `script`), `created_at`, `updated_at` — scripts and Python jobs the Executive saved to run again (`workflows.saved_tools`) |\n| `saved_tool_versions` | (`name`, `version`) PK, `description`, `script`, `tools` (JSON: the tools the script called when saved), `origin` (`chat` or `workflow:<name>/<step>`), `created_at` |\n| `saved_tool_runs` | `id`, `name`, `version`, `ok`, `calls`, `duration_ms`, `origin`, `at` — the newest 200 runs per tool |\n| `chat_messages` | `id`, `session_id`, `role`, `content`, `created_at`, `action_chips`, `stopped`, `sender_person_id`, `feedback`, `feedback_note`, `feedback_by_person_id`, `sources` (JSON `{sources: [{kind, title, url}], unavailable: [area]}` — what an assistant reply looked at and which areas it left out; NULL on older rows) |\n| `attunement_usage` | `day` (PK, UTC date), `calls` — the per-day ceiling on open-loop extraction model calls |\n| `attunement_profiles` | `person_id` (PK), `rules` (JSON: up to 4 `{text, basis, evidence}`), `locked`, `updated_at`, `updated_by`, `last_pass_at`, `pass_day`, `passes_today`, `last_message_id` — per-person working style plus pass pacing |\n| `attunement_profile_history` | `id`, `person_id`, `created_at`, `rules`, `locked`, `updated_by` — every change to a working style |\n| `proactive_outcomes` | `id`, `created_at`, `person_id`, `source`, `ref`, `channel`, `channel_ref`, `outbound_context_id`, `outcome` (`replied` / `acted` / `ignored` / `void`, NULL while open), `resolved_at` — one row per proactive DM (Attunement outcome ledger) |\n| `delegation_settings` | `person_id` (PK), `enabled` (default 0), `updated_at`, `updated_by` — Act as me, per person; no row means off |\n| `history_notes` | `id`, `person_id`, `visibility` (`private`), `source` (`approved_reply`), `channel` (`email`), `conversation_key` (a hash of the channel and thread, never the thread id), `counterpart`, `subject`, `kind` (`promised` / `agreed` / `declined` / `answered` / `asked` / `shared`), `summary`, `quote` (the person's own words), `due_date`, `trust`, `occurred_at`, `created_at`, `expires_at` (NULL: until forgotten), `pinned`, `correction`, `corrected_at`, `kept_from` (when the retention clock started: NULL means `occurred_at`; an unpin restarts it) — Always in the loop, one row per note |\n| `history_settings` | `id` (1), `retention_days` (NULL: until forgotten; no row means 90), `updated_at`, `updated_by` — the company retention |\n| `history_person_settings` | `person_id` (PK), `reply_notes` (default 0), `retention_days` (NULL: the company's), `updated_at`, `updated_by` |\n| `history_excluded` | `person_id`, `conversation_key` (PK together), `created_at` — conversations a person asked not to remember |\n| `history_chat_passes` | `person_id`, `day` (PK together, UTC `YYYY-MM-DD`), `passes` — note passes run on a person's chat messages that day, against the daily cap; older days pruned on write |\n| `delegation_voice` | `person_id` (PK), `profile` (JSON: `greetings` per audience, `sign_off`, `signature`, `length`, `formality`, up to 8 `habits`, up to 6 `avoid`, up to 3 `exemplars`), `locked`, `learned_at`, `sample_count`, `updated_at`, `updated_by` — \"How I write\" |\n| `delegation_voice_history` | `id`, `person_id`, `created_at`, `profile`, `locked`, `updated_by` — every change to a writing profile |\n| `delegation_drafts` | `id`, `person_id`, `source` (`chat` \\| `inbox`), `thread_id`, `draft_id`, `message_id`, `sent_message_id`, `created_at` — one row of ids per draft Act as me saved in someone's Gmail (no address, subject or text): the daily limit counts it, and \"How I write\" skips a chat draft's thread and a sent draft's message |\n| `delegation_inbox_watch` | `person_id` (PK), `enabled`, `watch_since` (reset on every off→on), `last_poll_at`, `status`, `backoff_until`, `failures`, `mailbox_drafts` (also save each reply in their Drafts; default 0, the reply lives on its card alone), `updated_at`, `updated_by` — **Draft replies to my inbox**, per person; no row means off |\n| `delegation_inbox_messages` | `id`, `person_id`, `message_id`, `thread_id`, `received_at`, `sender_key` (a hash of the address, never the address), `relation`, `outcome` (`processing`, `retry` — left for a later scan, `skipped`, `not_needed`, `deferred`, `failed`, `drafted`, `sent`, `dismissed`, `closed`, `expired`), `reason`, `classified`, `attempts`, `draft_id`, `decision_id`, `flags` (JSON), `created_at`, `updated_at`; `UNIQUE(person_id, message_id)` — the inbox watcher's cursor: what happened to each message it looked at |\n| `delegation_handle_it` | `person_id` (PK), `enabled`, `mode` (`careful`, `balanced` or `bold`), `levels` (the per-kind setting `mode` replaced, kept for older rows), `updated_at`, `updated_by` — **Handle it for me**, per person; no row means off. A `training` mode from before per-setting training moves to `balanced` with Replies and Follow-ups in training |\n| `delegation_training` | `person_id`, `setting` (`replies`, `follow_ups`, `actions` or `drafts`), `enabled`, `updated_at`, `updated_by`; primary key `(person_id, setting)` — Act as me settings in training, per person; no row means not in training |\n| `delegation_disclosure` | `person_id` (PK), `handle_it`, `approved_sends` (each NULL = as the install's `AI_DISCLOSURE_HANDLE_IT` / `AI_DISCLOSURE_APPROVED_SENDS`), `updated_at`, `updated_by` — the AI line on replies sent as that person; no row means the install's defaults |\n| `delegation_handled` | `id`, `person_id`, `thread_id`, `decision_id`, `sent_message_id`, `sent_at`; index on `(person_id, sent_at)` — one row per reply or follow-up Handle it for me sent on its own, read for its one-per-thread-a-day and daily limits |\n\n| `dynamic_workflows` | `name` (PK), `definition` (JSON of `DynamicWorkflowDef`), `is_active`, `created_at`, `updated_at`, `owner_person_id` (creator; set on insert only, kept out of the definition JSON) |\n| `workflow_approved_targets` | (`workflow_name`, `value`) PK, `key`, `approved_at`, `run_id` — targets a workflow's tool steps may write to without asking again; cleared when the workflow is deleted |\n| `executive_control` | `id` (PK, always 1), `paused`, `paused_at`, `paused_by`, `reason`, `updated_at` — the operator pause switch (`scheduler.pause`); operator-level, so client-slot switches carry it across |\n| `app_migrations` | `name` (PK), `applied_at` — one-shot data migrations (sweeps that must run exactly once per DB); schema DDL stays idempotent and is not recorded here |\n\n`department` and `session_id` are additive columns (default `''`); `session_id` scopes per-thread recall. `scheduled_actions.status` ∈ pending / running / done / failed / cancelled, and `scheduled_actions.kind` includes `dynamic_workflow` (a cadence-fired user-created workflow run) and `open_loop` (an Attunement open loop: `assigned_to_person_id` is the owner, `awaiting_response_since` the due time — cleared when the loop closes — and `scope_key` `loop:{owner}:{hash}`, unique among open loops via the partial index `idx_scheduled_open_loop_scope`). `chat_messages.sender_person_id` is the resolved rostered sender of a user message (never the session owner's principal fallback; NULL on legacy rows and for unrostered senders); `feedback` is `up` / `down` on an assistant reply. `dynamic_workflows` stores user-created (dynamic) workflow definitions interpreted by the generic `DynamicWorkflow` engine; the JSON `definition` holds free-text input fields and an ordered step list (specialist / approval_gate / synthesis) plus an optional cadence.\n\n### Legacy tables (talent / staff onboarding)\n\nThe talent (executive-search) and staff-onboarding features were removed. Their tables —\n`engagements`, `candidates`, `offers`, `onboarding_templates`, `onboarding_plans`,\n`onboarding_tasks` — are **not dropped**: no code creates, reads or writes them any more,\nbut rows survive on upgraded installs. `clients.slots` still wipes them when a blank client\nslot is activated (existence-guarded, a no-op on fresh installs) so a slot can never inherit\na previous client's candidate data.\n\n### Department / Goal / ChecklistItem\n\n| Model | Fields |\n|---|---|\n| `DepartmentConfig` | `slug`, `title`, `specialist_key`, `charter`, `authority_level`, `head_person_id`, `head_persona_slug`, `cadences`, `slack_channel_id`, `discord_channel_id`, `telegram_chat_id` |\n| `Goal` | `id`, `department_slug`, `period_type`, `period_value`, `key_result`, `target`, `current`, `status`, `created_at`, `updated_at`, `last_reviewed_at` |\n| `DepartmentState` | `config`, `goals`, `headcount`, `budget_usd`, `member_person_ids`, `updated_at` |\n\n`authority_level` ∈ `auto_execute` / `propose_only` / `escalate`; `Goal.status` ∈ `on_track` / `at_risk` / `off_track`.\n\n### Person\n\n| Field group | Fields |\n|---|---|\n| Identity | `id`, `full_name`, `role`, `is_principal`, `kind`, `department_slugs` |\n| Contact | `email`, `email_aliases`, `slack_user_id`, `telegram_chat_id`, `discord_user_id`, `preferred_channel` |\n| Routing | `availability`, `authority_scope`, `response_sla_hours`, `on_leave_until`, `reports_to_person_id` |\n| Lifecycle | `archived`, `created_at`, `updated_at` |\n\nExactly one `Person` should have `is_principal=True` — the fallback approver when no delegated person matches.\n\n`kind` is `team` (the default) or `contact` — someone outside the team, such as a client or contractor. It is an additive column, `kind TEXT NOT NULL DEFAULT 'team'`, added by the same `PRAGMA table_info`-guarded migration as `discord_user_id`, so every existing row is a team member. The principal is always `team`: the store refuses to write a principal contact (`PrincipalContactError`), and `initialize_db` repairs a hand-edited one on boot. A stored value the code does not know reads as `contact`. Every roster read in `people.store` and `people.registry` returns team members only unless it passes `include_contacts=True` (see Org).\n\n`email` is the one address a person signs in with; `email_aliases` (read from the `person_emails` table, up to 10) are other addresses they write from — they match inbound mail and may be emailed, but never sign in (see Org).\n\n### People-side tables\n\nThese live next to `people` (created by `people.store.initialize_db`, which also runs `people.roster_requests.initialize_tables`):\n\n| Table | Key columns |\n|---|---|\n| `person_emails` | `id`, `person_id` (FK → `people`, `ON DELETE CASCADE`), `email`, `email_norm` (trimmed + lowercased, **UNIQUE** across the roster), `source` (`manual` / `chat_tool` / `roster_request`), `roster_request_id`, `created_at` — a person's aliases; deleted when they are archived |\n| `roster_requests` | `id`, `channel` (`email` / `slack` / `discord` / `telegram`), `channel_ref` (the address or chat id), `display_name` (sanitised, unverified), `profile_email`, `on_company_domain`, `suggested_kind`, `suggested_person_id`, `status` (`pending` / `resolving` / `approved` / `linked` / `declined` / `expired` / `superseded`), `resolved_person_id`, `resolved_kind`, `resolved_via` (`web` / `slack` / `discord` / `telegram` / `email` / `people_page`), `resolved_at`, `token_hash` (the email answer token, hashed; cleared on resolve), `confirm_message_id`, `notified_at`, `notified_channel`, `ack_sent_at`, `alert_id`, `message_count`, `first_seen_at`, `last_seen_at`, `expires_at` — one pending row per `(channel, channel_ref)` (partial unique index) |\n| `roster_request_messages` | `id`, `request_id` (FK, cascade), `external_id` (unique per request, so a redelivery is held once), `payload_json` (what the adapter needs to replay it), `preview`, `received_at`, `replay_status` (`held` / `replaying` / `replayed` / `dropped` / `failed` / `unavailable`), `replayed_at` — at most 10 per request; `payload_json` and `preview` are blanked once a message is replayed or dropped |\n| `roster_ack_log` | `id`, `channel`, `channel_ref`, `sent_at` — one row per acknowledgement sent to a new sender, for the per-sender window and the daily cap |\n\n**Take the lead tables** (`orchestrator.take_the_lead`, created on first use, per company in `clients.slots` and `reset_all_state`): `take_the_lead` (`scope` PK: `executive` or `person:<id>`, `enabled`, `training` (added in place on older databases), `updated_at`, `updated_by`; each kind's Ask first is a `decision_class_state` row for `take_the_lead_action:<kind>`); `take_the_lead_rules` (`id`, `scope`: `company` or `person:<id>`, `kind` person|domain|words|amount, `value`, `created_by`, `created_at`); `take_the_lead_log` (`id`, `at`, `scope`, `source` reflection|research|scheduled, `tool`, `summary`, `status` done|waiting|approved|declined|failed, `why`, `decision_id`). A held action is a `decision_instances` row of class `take_the_lead_action` whose payload carries the tool, its input, whether it is an MCP tool, the gate kind or rule, the reason and the summary. `take_the_lead_allowed` (`id`, `key` unique: `<feature>|<action>|<who>`, `label`, `feature`, `example` (JSON of the edited fields, or empty), `uses`, `last_used_at`, `created_by`, `created_at`, `decision_id`, `removed_at`: set by Remove, so only approvals after it count toward a new suggestion, `person_id`: whose it is for a feature that learns per person, Act as me's `act_as_me|person:<id>|<reply|follow_up|message|invite|style>|<who>`, NULL for Take the lead; added in place on older databases): what it may do on its own in training, at most `ALLOWED_MAX` (200) per feature and person.",
"mermaid": null,
- "generated_at": "2026-10-10T16:26:53Z"
+ "generated_at": "2026-10-13T15:30:00Z"
}
diff --git a/packages/core/openexecutive/audit/logger.py b/packages/core/openexecutive/audit/logger.py
index 9db8141f..66047f29 100644
--- a/packages/core/openexecutive/audit/logger.py
+++ b/packages/core/openexecutive/audit/logger.py
@@ -99,6 +99,7 @@ EVENT_TYPES: tuple[str, ...] = (
"delegation_actions_handled", # a card of actions they allowed was carried out on its own (in training)
"delegation_actions_held", # a card of allowed actions waited for them after all; details.reason says why
"delegation_training_changed", # a person put Act as me settings in or out of training (PUT /delegation/training)
+ "delegation_disclosure_changed", # a person changed the AI line on replies sent as them (PUT /delegation/disclosure)
"fact_retired", # the principal (or the teammate who recorded it) retired a standing fact from the Pulse page (memory/facts.py)
"fact_reviewed", # the principal approved or declined a teammate's proposed standing fact
"fact_approval_changed", # the principal turned "needs my approval" on or off for a teammate's standing facts
diff --git a/packages/core/openexecutive/config.py b/packages/core/openexecutive/config.py
index f30b1fe7..17651587 100644
--- a/packages/core/openexecutive/config.py
+++ b/packages/core/openexecutive/config.py
@@ -490,6 +490,18 @@ class Settings(BaseSettings):
# identity addendum so the model has a concrete self-name and never
# falls back to signing as a person from the company People roster.
exec_display_name: str = Field("Open Executive", alias="EXEC_DISPLAY_NAME")
+ # AI disclosure (orchestrator.disclosure): a line on the mail and chat
+ # messages the Executive sends saying it is an AI assistant. On unless
+ # turned off. The texts replace the built-in wording; one that doesn't
+ # say "AI" is ignored, so the line can't be reworded into hiding it.
+ ai_disclosure: bool = Field(True, alias="AI_DISCLOSURE")
+ ai_disclosure_mail_text: str = Field("", alias="AI_DISCLOSURE_MAIL_TEXT", max_length=500)
+ ai_disclosure_chat_text: str = Field("", alias="AI_DISCLOSURE_CHAT_TEXT", max_length=200)
+ # Where each person's own switches start (delegation.disclosure): a line
+ # on the replies Handle it for me sends as them on its own, and on the
+ # drafts they approve and send themselves. Each person can change theirs.
+ ai_disclosure_handle_it: bool = Field(True, alias="AI_DISCLOSURE_HANDLE_IT")
+ ai_disclosure_approved_sends: bool = Field(False, alias="AI_DISCLOSURE_APPROVED_SENDS")
email_poll_interval_seconds: int = Field(60, alias="EMAIL_POLL_INTERVAL_SECONDS")
# Which Google services the co-located workspace-mcp child loads, and at
diff --git a/packages/core/openexecutive/delegation/disclosure.py b/packages/core/openexecutive/delegation/disclosure.py
new file mode 100644
index 00000000..049554e4
--- /dev/null
+++ b/packages/core/openexecutive/delegation/disclosure.py
@@ -0,0 +1,137 @@
+"""Act as me: the AI disclosure line on replies sent as a person.
+
+Mail the Executive sends as itself carries ``orchestrator.disclosure``'s
+line. Replies the inbox watcher writes in a person's own voice go out as
+*them*, so each person decides for their own mailbox, with two switches:
+
+- **Handle it** (on unless the install says otherwise): a reply Handle it
+ for me will send on its own, which nobody reads before it goes, ends with
+ ``LINE``.
+- **Approved sends** (off unless the install says otherwise): a draft the
+ person reads and sends with Send ends with it too.
+
+Both start from the install's ``AI_DISCLOSURE_HANDLE_IT`` /
+``AI_DISCLOSURE_APPROVED_SENDS`` until the person changes theirs (one row in
+``delegation_disclosure``, NULL meaning "as the install says"). The line is
+written into the reply when it is drafted (``sign``), so the card, the
+mailbox draft and what is sent are the same words, and Send's checks see
+them; a switch changed later applies to the next reply. ``AI_DISCLOSURE``
+off turns both off. Unreadable settings fall back to the install's.
+"""
+from __future__ import annotations
+
+import logging
+import sqlite3
+from dataclasses import dataclass
+from datetime import UTC, datetime
+from pathlib import Path
+
+from openexecutive.delegation.schema import DISCLOSURE_TABLE, ensure_schema
+
+logger = logging.getLogger(__name__)
+
+LINE = "Drafted and sent by my AI assistant."
+
+
+@dataclass(frozen=True)
+class Disclosure:
+ person_id: int
+ # Effective: the person's own choice, else the install's.
+ handle_it: bool
+ approved_sends: bool
+
+
+def _defaults() -> tuple[bool, bool]:
+ from openexecutive.config import get_settings
+
+ settings = get_settings()
+ return bool(settings.ai_disclosure_handle_it), bool(settings.ai_disclosure_approved_sends)
+
+
+def _db(db_path: Path | None) -> Path:
+ if db_path is not None:
+ return db_path
+ from openexecutive.memory import episodic
+
+ return Path(episodic.DB_PATH)
+
+
+def _connect(db_path: Path | None) -> sqlite3.Connection:
+ conn = sqlite3.connect(str(_db(db_path)))
+ conn.row_factory = sqlite3.Row
+ ensure_schema(conn)
+ return conn
+
+
+def get(person_id: int, *, db_path: Path | None = None) -> Disclosure:
+ """``person_id``'s two switches. Never raises."""
+ handle_it, approved = _defaults()
+ try:
+ conn = _connect(db_path)
+ try:
+ row = conn.execute(
+ f"SELECT handle_it, approved_sends FROM {DISCLOSURE_TABLE} WHERE person_id = ?", # noqa: S608
+ (person_id,),
+ ).fetchone()
+ finally:
+ conn.close()
+ except Exception:
+ logger.warning("delegation.disclosure: couldn't read the switches — using the install's", exc_info=True)
+ row = None
+ if row is not None:
+ if row["handle_it"] is not None:
+ handle_it = bool(row["handle_it"])
+ if row["approved_sends"] is not None:
+ approved = bool(row["approved_sends"])
+ return Disclosure(person_id=person_id, handle_it=handle_it, approved_sends=approved)
+
+
+def set_(
+ person_id: int,
+ *,
+ handle_it: bool | None = None,
+ approved_sends: bool | None = None,
+ updated_by: str,
+ db_path: Path | None = None,
+) -> Disclosure:
+ """Change ``person_id``'s own switches (callers authorize first); None
+ leaves one as it is, still following the install if it did."""
+ conn = _connect(db_path)
+ try:
+ conn.execute(
+ f"INSERT INTO {DISCLOSURE_TABLE} (person_id, handle_it, approved_sends, updated_at, updated_by) " # noqa: S608
+ "VALUES (?, ?, ?, ?, ?) ON CONFLICT(person_id) DO UPDATE SET "
+ "handle_it = COALESCE(excluded.handle_it, handle_it), "
+ "approved_sends = COALESCE(excluded.approved_sends, approved_sends), "
+ "updated_at = excluded.updated_at, updated_by = excluded.updated_by",
+ (
+ person_id,
+ None if handle_it is None else int(handle_it),
+ None if approved_sends is None else int(approved_sends),
+ datetime.now(UTC).isoformat(),
+ updated_by,
+ ),
+ )
+ conn.commit()
+ finally:
+ conn.close()
+ return get(person_id, db_path=db_path)
+
+
+def wanted(person_id: int, *, on_its_own: bool) -> bool:
+ """Whether a reply drafted for ``person_id`` gets the line: one Handle it
+ will send on its own, or one waiting for them to send."""
+ from openexecutive.orchestrator.disclosure import enabled
+
+ if not enabled():
+ return False
+ switches = get(person_id)
+ return switches.handle_it if on_its_own else switches.approved_sends
+
+
+def sign(person_id: int, body: str, *, on_its_own: bool) -> str:
+ """``body`` as the reply should be drafted: with ``LINE`` at the end
+ when ``person_id``'s switch for this kind of send is on."""
+ if not body.strip() or LINE in body or not wanted(person_id, on_its_own=on_its_own):
+ return body
+ return f"{body.rstrip()}\n\n{LINE}"
diff --git a/packages/core/openexecutive/delegation/follow_ups.py b/packages/core/openexecutive/delegation/follow_ups.py
index 5b0e3ec6..5b2bfdcb 100644
--- a/packages/core/openexecutive/delegation/follow_ups.py
+++ b/packages/core/openexecutive/delegation/follow_ups.py
@@ -130,6 +130,7 @@ async def _follow_up(
was made."""
from openexecutive.config import get_settings
from openexecutive.delegation import caps, card_drafts, drafts, handle_it, inbox
+ from openexecutive.delegation import disclosure as ai_disclosure
from openexecutive.delegation.ghostwriter import ComposeError, Recipient, compose
from openexecutive.delegation.gmail import DraftSpec, references_header
from openexecutive.delegation.inbox_classifier import Verdict
@@ -184,6 +185,7 @@ async def _follow_up(
held = handle_it.follow_up_refusal(
person.id, sent, reply, relation=relation, own=own, exec_address=exec_address, now=now,
)
+ reply.body = ai_disclosure.sign(person.id, reply.body, on_its_own=held is None)
spec = DraftSpec(
to=reply.to, cc=reply.cc, subject=reply.subject, body=reply.body, thread_id=thread.id,
in_reply_to=reply.in_reply_to, references=reply.references,
diff --git a/packages/core/openexecutive/delegation/inbox.py b/packages/core/openexecutive/delegation/inbox.py
index 3fdb31a8..8b07ea39 100644
--- a/packages/core/openexecutive/delegation/inbox.py
+++ b/packages/core/openexecutive/delegation/inbox.py
@@ -1095,6 +1095,7 @@ async def _consider(
(``_retry_later``); a Gmail failure is raised for the scan to back off."""
from openexecutive.config import get_settings
from openexecutive.delegation import caps, card_drafts, drafts, handle_it
+ from openexecutive.delegation import disclosure as ai_disclosure
from openexecutive.delegation.ghostwriter import ComposeError
from openexecutive.delegation.gmail import DraftSpec, GmailError
from openexecutive.delegation.inbox_classifier import addressing, classify, wants_draft
@@ -1166,6 +1167,9 @@ async def _consider(
person.id, message, thread, reply, verdict, relation=relation, own=own,
exec_address=exec_address, now=now,
) if handling.enabled else "level"
+ # The AI line, by the person's own switch for this kind of send: in
+ # the words themselves, so card, draft and what goes all match.
+ reply.body = ai_disclosure.sign(person.id, reply.body, on_its_own=held is None)
names = (person.full_name or "").split()
# Write from the address the mail went to: one of their send-as
# addresses rather than the primary when only that one was used.
diff --git a/packages/core/openexecutive/delegation/schema.py b/packages/core/openexecutive/delegation/schema.py
index f39defd2..2ca73d55 100644
--- a/packages/core/openexecutive/delegation/schema.py
+++ b/packages/core/openexecutive/delegation/schema.py
@@ -21,6 +21,7 @@ HANDLED_TABLE = "delegation_handled"
REMINDERS_TABLE = "delegation_reminders"
MAIL_READ_TABLE = "delegation_mail_read"
TRAINING_TABLE = "delegation_training"
+DISCLOSURE_TABLE = "delegation_disclosure"
TABLES: tuple[str, ...] = (
SETTINGS_TABLE,
@@ -35,6 +36,7 @@ TABLES: tuple[str, ...] = (
REMINDERS_TABLE,
MAIL_READ_TABLE,
TRAINING_TABLE,
+ DISCLOSURE_TABLE,
)
_DDL: tuple[str, ...] = (
@@ -192,6 +194,16 @@ _DDL: tuple[str, ...] = (
" updated_by TEXT,"
" PRIMARY KEY (person_id, setting)"
")",
+ # The AI disclosure line on what is sent as a person (delegation.disclosure):
+ # one row per person who changed theirs. NULL follows the install's
+ # AI_DISCLOSURE_HANDLE_IT / AI_DISCLOSURE_APPROVED_SENDS.
+ f"CREATE TABLE IF NOT EXISTS {DISCLOSURE_TABLE} ("
+ " person_id INTEGER PRIMARY KEY,"
+ " handle_it INTEGER,"
+ " approved_sends INTEGER,"
+ " updated_at TEXT,"
+ " updated_by TEXT"
+ ")",
)
diff --git a/packages/core/openexecutive/integrations/discord_bot.py b/packages/core/openexecutive/integrations/discord_bot.py
index 508d8b57..670d05a4 100644
--- a/packages/core/openexecutive/integrations/discord_bot.py
+++ b/packages/core/openexecutive/integrations/discord_bot.py
@@ -1016,7 +1016,9 @@ async def _handle_message(
# to silently no-op than to spam the channel with an error
# message that itself confuses users.
if response.strip():
- for chunk in _split_message(response):
+ from openexecutive.orchestrator.disclosure import with_chat_line
+
+ for chunk in _split_message(with_chat_line(response, f"discord:{session_id}")):
await effective_send_fn(chunk)
else:
logger.info(
diff --git a/packages/core/openexecutive/integrations/google_chat.py b/packages/core/openexecutive/integrations/google_chat.py
index e0e9342d..10c5745d 100644
--- a/packages/core/openexecutive/integrations/google_chat.py
+++ b/packages/core/openexecutive/integrations/google_chat.py
@@ -333,8 +333,11 @@ async def _process_and_reply(
channel_context_block=build_channel_context_block("google_chat"),
attachment_blocks=image_blocks or None,
)
+ from openexecutive.orchestrator.disclosure import with_chat_line
+
await asyncio.to_thread(
- send_reply, space_name, thread_name, response, service_account_file, service_account_email
+ send_reply, space_name, thread_name, with_chat_line(response, session_id),
+ service_account_file, service_account_email,
)
except Exception:
logger.exception("Google Chat: handler error for message %s", message_name)
diff --git a/packages/core/openexecutive/integrations/slack_bot.py b/packages/core/openexecutive/integrations/slack_bot.py
index a2f8b5b9..0b8244b0 100644
--- a/packages/core/openexecutive/integrations/slack_bot.py
+++ b/packages/core/openexecutive/integrations/slack_bot.py
@@ -1032,7 +1032,12 @@ async def create_slack_app():
# No unfurls: Slack would fetch a link in the reply on its own,
# and a reply can quote mail a sender wrote to carry data out in one.
- await say(text=response, thread_ts=thread_ts, unfurl_links=False, unfurl_media=False)
+ from openexecutive.orchestrator.disclosure import with_chat_line
+
+ await say(
+ text=with_chat_line(response, f"slack:{session_id}"),
+ thread_ts=thread_ts, unfurl_links=False, unfurl_media=False,
+ )
# Persist AFTER the reply lands, and never let a persistence
# failure trigger the user-facing error path — the user
diff --git a/packages/core/openexecutive/integrations/telegram_bot.py b/packages/core/openexecutive/integrations/telegram_bot.py
index 996c3a7c..49a9b35d 100644
--- a/packages/core/openexecutive/integrations/telegram_bot.py
+++ b/packages/core/openexecutive/integrations/telegram_bot.py
@@ -351,7 +351,9 @@ async def _process_and_reply(
channel_context_block=build_channel_context_block("telegram"),
person_id=person_id,
)
- await send_message(token, chat_id, response)
+ from openexecutive.orchestrator.disclosure import with_chat_line
+
+ await send_message(token, chat_id, with_chat_line(response, f"telegram:{session_id}"))
except Exception:
logger.exception("Telegram: handler error for message %s", message_id)
try:
diff --git a/packages/core/openexecutive/orchestrator/broadcast_tools.py b/packages/core/openexecutive/orchestrator/broadcast_tools.py
index c10a40b0..b18a2f09 100644
--- a/packages/core/openexecutive/orchestrator/broadcast_tools.py
+++ b/packages/core/openexecutive/orchestrator/broadcast_tools.py
@@ -243,6 +243,9 @@ async def _post_telegram_chat(chat_id: str, text: str) -> dict[str, Any]:
async def _dispatch(
integration: str, channel_id: str, text: str
) -> dict[str, Any]:
+ from openexecutive.orchestrator.disclosure import with_chat_line
+
+ text = with_chat_line(text, f"{integration}:{channel_id}")
if integration == "slack":
return await _post_slack_channel(channel_id, text)
if integration == "discord":
diff --git a/packages/core/openexecutive/orchestrator/disclosure.py b/packages/core/openexecutive/orchestrator/disclosure.py
new file mode 100644
index 00000000..97647707
--- /dev/null
+++ b/packages/core/openexecutive/orchestrator/disclosure.py
@@ -0,0 +1,213 @@
+"""AI disclosure: a line on what the Executive sends saying it is an AI.
+
+People the Executive writes to should be able to tell they are hearing from
+an AI assistant. Plain code adds the line, never the prompt, so a reply the
+model writes can't leave it out:
+
+- **Mail** from the Executive's own mailbox gets ``mail_line()`` below the
+ signature: the MCP gateway adds it to every Gmail / Outlook send and draft
+ after the outbound gates pass (``with_mail_line``), which covers replies
+ the model writes and the fixed mails code sends (``send_from_executive``).
+ Mail only to the principal goes without it: they know.
+- **Chat** (Slack, Telegram, Discord, Google Chat) gets the shorter
+ ``chat_line()`` on the first message in a conversation, then again after
+ ``CHAT_REPEAT`` of quiet (``with_chat_line``), so a busy channel isn't
+ marked on every reply. What a conversation last saw is kept in memory, so
+ a restart shows it once more.
+
+Sends *as a person* (Act as me) are ``delegation.disclosure``'s.
+
+``AI_DISCLOSURE`` turns it all off. ``AI_DISCLOSURE_MAIL_TEXT`` /
+``AI_DISCLOSURE_CHAT_TEXT`` reword the lines; a text that doesn't say "AI"
+is ignored for the built-in one. Everything here fails open: a line that
+can't be built is left off, never a send refused.
+"""
+from __future__ import annotations
+
+import html
+import logging
+import re
+import threading
+import time
+
+logger = logging.getLogger(__name__)
+
+# Shown again in a conversation after this long without one.
+CHAT_REPEAT = 24 * 60 * 60.0
+# The most conversations remembered; the oldest go first.
+_MAX_CONVERSATIONS = 5000
+
+_SAYS_AI = re.compile(r"\bAI\b|artificial intelligence", re.IGNORECASE)
+_MAIL_RULE = "---"
+
+_lock = threading.Lock()
+_last_shown: dict[str, float] = {}
+
+
+def enabled() -> bool:
+ from openexecutive.config import get_settings
+
+ try:
+ return getattr(get_settings(), "ai_disclosure", True) is True
+ except Exception:
+ return True
+
+
+def says_ai(text: str) -> bool:
+ """Whether ``text`` tells the reader they are dealing with an AI."""
+ return bool(_SAYS_AI.search(text or ""))
+
+
+def _custom(text: str, setting: str) -> str | None:
+ text = " ".join((text or "").split())
+ if not text:
+ return None
+ if says_ai(text):
+ return text
+ logger.warning("disclosure: %s doesn't say it's an AI, so the built-in line is used", setting)
+ return None
+
+
+def _company_name() -> str:
+ try:
+ from openexecutive.onboarding.profile_builder import load_or_create_profile
+
+ return " ".join(load_or_create_profile().name.split())
+ except Exception:
+ return ""
+
+
+def _principal_email() -> str:
+ try:
+ from openexecutive.people.registry import get_principal
+
+ principal = get_principal()
+ return (principal.email or "").strip() if principal is not None else ""
+ except Exception:
+ return ""
+
+
+def mail_line() -> str:
+ """The line under the Executive's mail."""
+ from openexecutive.config import get_settings
+
+ settings = get_settings()
+ custom = _custom(settings.ai_disclosure_mail_text, "AI_DISCLOSURE_MAIL_TEXT")
+ if custom:
+ return custom
+ name = settings.exec_display_name.strip() or "This assistant"
+ company = _company_name()
+ line = f"{name} is an AI assistant working for {company}." if company else f"{name} is an AI assistant."
+ contact = _principal_email()
+ if contact:
+ line += f" To reach a person, reply and ask, or write to {contact}."
+ return line
+
+
+def chat_line() -> str:
+ """The shorter line on the Executive's chat messages."""
+ from openexecutive.config import get_settings
+
+ settings = get_settings()
+ custom = _custom(settings.ai_disclosure_chat_text, "AI_DISCLOSURE_CHAT_TEXT")
+ if custom:
+ return custom
+ return f"{settings.exec_display_name.strip() or 'Open Executive'} · AI assistant"
+
+
+# The line for mail to someone the roster never checked (the acknowledgement
+# to a new sender): no name, company or address, like the message itself.
+BARE_MAIL_LINE = "This message was sent by an AI assistant."
+
+# Markup that can keep whatever follows it from showing: an unclosed comment,
+# style or script, or text styled away. An HTML body with any of it isn't sent
+# (``hides_what_follows``), so the line can't be hidden after it.
+_HIDING = re.compile(
+ r"<!--|<style|<script|<template|<noscript|<iframe|<object|display\s*:\s*none|visibility\s*:\s*hidden"
+ r"|(?:font-size|opacity|height)\s*:\s*0(?![.\d])|mso-hide",
+ re.IGNORECASE,
+)
+
+
+def hides_what_follows(html_body: str) -> bool:
+ """Whether HTML mail carries markup that could hide the line after it."""
+ return bool(_HIDING.search(html_body or ""))
+
+
+def _ends_with_line(body: str, line: str, *, as_html: bool) -> bool:
+ """Whether ``body`` already ends with the line *as shown*: in a quoted
+ earlier message or a comment it doesn't count."""
+ if as_html:
+ from openexecutive.utils.html_tags import strip_tags
+
+ body = html.unescape(strip_tags(body, repl="\n"))
+ lines = body.rstrip().splitlines()
+ return bool(lines) and lines[-1].strip() == line
+
+
+def add_mail_line(body: str, *, as_html: bool = False, bare: bool = False) -> str:
+ """``body`` with the mail line below it, once."""
+ try:
+ line = BARE_MAIL_LINE if bare else mail_line()
+ except Exception:
+ logger.warning("disclosure: couldn't build the mail line", exc_info=True)
+ return body
+ if _ends_with_line(body, line, as_html=as_html):
+ return body
+ if as_html:
+ return f'{body}<hr><p style="color:#666;font-size:12px">{html.escape(line)}</p>'
+ return f"{body.rstrip()}\n\n{_MAIL_RULE}\n{line}"
+
+
+def with_mail_line(
+ body: str, *, recipients: list[str] | None = None, as_html: bool = False, bare: bool = False,
+) -> str:
+ """The Executive's mail body as it should go: with the line, unless
+ disclosure is off or it only goes to the principal. ``bare`` uses
+ ``BARE_MAIL_LINE``, for a recipient the roster never checked."""
+ if not isinstance(body, str) or not applies(recipients):
+ return body
+ return add_mail_line(body, as_html=as_html, bare=bare)
+
+
+def applies(recipients: list[str] | None = None) -> bool:
+ """Whether mail to ``recipients`` gets the line: disclosure is on and it
+ doesn't only go to the principal (unknown recipients count as others)."""
+ if not enabled():
+ return False
+ if recipients:
+ principal = _principal_email().lower()
+ if principal and all(r.strip().lower() == principal for r in recipients):
+ return False
+ return True
+
+
+def with_chat_line(text: str, conversation: str, *, now: float | None = None) -> str:
+ """``text`` as it should go to ``conversation`` (any stable key for one
+ chat, DM or thread): with the chat line when that conversation hasn't
+ had one for ``CHAT_REPEAT``."""
+ if not enabled() or not isinstance(text, str) or not text.strip():
+ return text
+ try:
+ line = chat_line()
+ except Exception:
+ logger.warning("disclosure: couldn't build the chat line", exc_info=True)
+ return text
+ if line in text:
+ return text
+ at = time.time() if now is None else now
+ with _lock:
+ last = _last_shown.get(conversation)
+ if last is not None and at - last < CHAT_REPEAT:
+ return text
+ _last_shown.pop(conversation, None)
+ _last_shown[conversation] = at
+ while len(_last_shown) > _MAX_CONVERSATIONS:
+ _last_shown.pop(next(iter(_last_shown)))
+ return f"{text.rstrip()}\n\n_{line}_"
+
+
+def reset() -> None:
+ """Forget which conversations have seen the chat line (tests)."""
+ with _lock:
+ _last_shown.clear()
diff --git a/packages/core/openexecutive/orchestrator/mcp_gateway.py b/packages/core/openexecutive/orchestrator/mcp_gateway.py
index 604bf7a3..a556f48c 100644
--- a/packages/core/openexecutive/orchestrator/mcp_gateway.py
+++ b/packages/core/openexecutive/orchestrator/mcp_gateway.py
@@ -1553,6 +1553,98 @@ def _check_calendar_attendees(tool: str, arguments: dict[str, Any]) -> str | Non
return None
+def _mail_recipients(*values: Any) -> list[str]:
+ """Every address in Gmail-style ``to`` / ``cc`` / ``bcc`` values (a
+ string or a list of them)."""
+ found: list[str] = []
+ for value in values:
+ items = value if isinstance(value, list) else [value]
+ found += [a for _, a in getaddresses([v for v in items if isinstance(v, str)]) if a]
+ return found
+
+
+def _m365_recipients(message: dict[str, Any]) -> list[str]:
+ found: list[str] = []
+ for key in ("toRecipients", "ccRecipients", "bccRecipients"):
+ entries = message.get(key)
+ for entry in entries if isinstance(entries, list) else []:
+ address = entry.get("emailAddress") if isinstance(entry, dict) else None
+ if isinstance(address, dict) and isinstance(address.get("address"), str):
+ found.append(address["address"])
+ return found
+
+
+def _with_disclosure(
+ tool_name: Any, normalized: str, arguments: dict[str, Any], *, bare: bool = False,
+) -> dict[str, Any] | str:
+ """The Executive's own mail send or draft with the AI disclosure line
+ under its body (``orchestrator.disclosure``), or a JSON refusal for HTML
+ that could hide it. Runs after every gate, on a copy; a shape it doesn't
+ know is left as it is. ``bare``: the acknowledgement to a new sender,
+ whose line names nobody."""
+ from openexecutive.orchestrator.disclosure import applies, hides_what_follows, with_mail_line
+
+ def html_refusal(key: str, text: str, recipients: list[str] | None) -> str | None:
+ if applies(recipients) and hides_what_follows(text):
+ return _refuse(
+ str(tool_name), key, "<hidden-markup>",
+ reason=(
+ "this HTML has markup that can hide text after it (a comment, a style or "
+ "script block, or text styled out of sight), so the line saying it was "
+ "written by an AI could be hidden — refusing. Send it as plain text, or "
+ "as HTML without that markup."
+ ),
+ )
+ return None
+
+ if tool_name in _GATED_GMAIL_TOOLS:
+ out = dict(arguments)
+ recipients = _mail_recipients(out.get("to"), out.get("cc"), out.get("bcc"))
+ if isinstance(out.get("html_body"), str) and out["html_body"]:
+ refused = html_refusal("html_body", out["html_body"], recipients)
+ if refused is not None:
+ return refused
+ out["html_body"] = with_mail_line(out["html_body"], recipients=recipients, as_html=True, bare=bare)
+ if isinstance(out.get("body"), str) and out["body"]:
+ as_html = str(out.get("body_format") or "").lower() == "html"
+ refused = html_refusal("body", out["body"], recipients) if as_html else None
+ if refused is not None:
+ return refused
+ out["body"] = with_mail_line(out["body"], recipients=recipients, as_html=as_html, bare=bare)
+ return out
+ if normalized not in _GATED_M365_MAIL_TOOLS or normalized == _M365_SEND_DRAFT_TOOL:
+ return arguments
+ body = arguments.get("body")
+ if not isinstance(body, dict):
+ return arguments
+ body = dict(body)
+ # A reply or forward carries its words as Comment, which Graph puts in
+ # the HTML reply; a new message or a draft as a Graph message (under
+ # Message for send_mail).
+ if isinstance(body.get("Comment"), str) and body["Comment"]:
+ refused = html_refusal("Comment", body["Comment"], None)
+ if refused is not None:
+ return refused
+ body["Comment"] = with_mail_line(body["Comment"], as_html=True)
+ return {**arguments, "body": body}
+ wrapped = isinstance(body.get("Message"), dict)
+ message = body["Message"] if wrapped else body
+ content = message.get("body")
+ if not (isinstance(content, dict) and isinstance(content.get("content"), str) and content["content"]):
+ return arguments
+ as_html = str(content.get("contentType") or "").lower() == "html"
+ recipients = _m365_recipients(message)
+ refused = html_refusal("body", content["content"], recipients) if as_html else None
+ if refused is not None:
+ return refused
+ content = {
+ **content,
+ "content": with_mail_line(content["content"], recipients=recipients, as_html=as_html),
+ }
+ message = {**message, "body": content}
+ return {**arguments, "body": {**body, "Message": message} if wrapped else message}
+
+
def _pin_calendar_notifications(arguments: dict[str, Any]) -> dict[str, Any]:
"""Keep a manage_event call that named no attendees from emailing the
event's existing guests.
@@ -3030,9 +3122,22 @@ class MCPGateway:
if isinstance(expanded, str):
return expanded
arguments, attached_artifacts = expanded
+ # Last, once every gate passed: the AI disclosure line, on what goes
+ # out only; what is recorded below stays the words the model wrote.
+ wire_arguments = arguments
+ if tool_name in _GATED_GMAIL_TOOLS or normalized in _GATED_M365_MAIL_TOOLS:
+ # The roster acknowledgement (the one send to an address nobody
+ # checked) gets the line that names nobody, as the ack itself.
+ grant = _roster_ack.get()
+ disclosed = _with_disclosure(
+ tool_name, normalized, arguments, bare=bool(grant is not None and grant["used"]),
+ )
+ if isinstance(disclosed, str):
+ return disclosed
+ wire_arguments = disclosed
result = await session.call_tool(
"call_tool",
- {"tool_name": tool_input["name"], "arguments": arguments},
+ {"tool_name": tool_input["name"], "arguments": wire_arguments},
)
result_text = result.content[0].text if result.content else json.dumps({"result": None})
# A pinned tool called before startup priming reached it: the prompt
@@ -3042,7 +3147,7 @@ class MCPGateway:
if _is_undiscovered_refusal(tool_name, result_text) and await self._discover(tool_name):
result = await session.call_tool(
"call_tool",
- {"tool_name": tool_input["name"], "arguments": arguments},
+ {"tool_name": tool_input["name"], "arguments": wire_arguments},
)
result_text = result.content[0].text if result.content else json.dumps({"result": None})
# On the final result, retry or not: roster answer tokens are hidden
diff --git a/packages/core/openexecutive/orchestrator/schedule_tools.py b/packages/core/openexecutive/orchestrator/schedule_tools.py
index 9b14343e..f5e6786c 100644
--- a/packages/core/openexecutive/orchestrator/schedule_tools.py
+++ b/packages/core/openexecutive/orchestrator/schedule_tools.py
@@ -27,6 +27,8 @@ from collections.abc import Awaitable, Callable, Iterator
from datetime import UTC, datetime, timedelta
from typing import Any
+from openexecutive.orchestrator.disclosure import with_chat_line
+
logger = logging.getLogger(__name__)
@@ -892,7 +894,9 @@ async def handle_send_telegram_message(tool_input: dict[str, Any]) -> str:
from openexecutive.audit import log_event as audit_log
try:
- msg_id = await send_message(settings.telegram_bot_token, chat_id, text)
+ msg_id = await send_message(
+ settings.telegram_bot_token, chat_id, with_chat_line(text, f"telegram:{chat_id}")
+ )
except Exception as exc:
logger.exception("send_telegram_message: send failed")
audit_log(
@@ -985,7 +989,9 @@ async def handle_send_slack_dm(tool_input: dict[str, Any]) -> str:
client = AsyncWebClient(token=settings.slack_bot_token)
try:
# No previews: Slack would fetch a link in the text to unfurl it.
- result = await client.chat_postMessage(channel=user_id, text=text, unfurl_links=False, unfurl_media=False)
+ result = await client.chat_postMessage(
+ channel=user_id, text=with_chat_line(text, f"slack_dm:{user_id}"), unfurl_links=False, unfurl_media=False,
+ )
except Exception as exc:
logger.exception("send_slack_dm: send failed")
return json.dumps({"error": f"send failed: {exc}"})
@@ -1128,7 +1134,7 @@ async def handle_send_discord_dm(tool_input: dict[str, Any]) -> str:
from openexecutive.audit import log_event as audit_log
try:
- msg_id = await send_dm(discord_user_id, text)
+ msg_id = await send_dm(discord_user_id, with_chat_line(text, f"discord_dm:{discord_user_id}"))
except Exception as exc:
logger.exception("send_discord_dm: send failed")
audit_log(
diff --git a/packages/core/tests/conftest.py b/packages/core/tests/conftest.py
index 03857d1b..17cb6d7b 100644
--- a/packages/core/tests/conftest.py
+++ b/packages/core/tests/conftest.py
@@ -46,6 +46,19 @@ from openexecutive.config import Settings # noqa: E402
Settings.model_config["env_file"] = None
[email protected](autouse=True)
+def _no_ai_disclosure(monkeypatch: pytest.MonkeyPatch):
+ """Leave the AI disclosure line (orchestrator.disclosure) off what tests
+ send, so a test's exact-text assertions don't depend on it, and forget
+ which chats have seen it. test_ai_disclosure.py turns it back on."""
+ from openexecutive.orchestrator import disclosure
+
+ monkeypatch.setenv("AI_DISCLOSURE", "false")
+ disclosure.reset()
+ yield
+ disclosure.reset()
+
+
@pytest.fixture(autouse=True)
def reset_active_gateway():
"""Ensure the module-level MCP gateway singleton is cleared between tests."""
diff --git a/packages/core/tests/unit/test_ai_disclosure.py b/packages/core/tests/unit/test_ai_disclosure.py
new file mode 100644
index 00000000..b184c348
--- /dev/null
+++ b/packages/core/tests/unit/test_ai_disclosure.py
@@ -0,0 +1,257 @@
+"""AI disclosure (orchestrator/disclosure.py, delegation/disclosure.py): the
+line on what the Executive sends, and on replies sent as a person."""
+from __future__ import annotations
+
+from collections.abc import Iterator
+from pathlib import Path
+from typing import Any
+
+import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
+
+from openexecutive.api.routes import delegation as route
+from openexecutive.delegation import disclosure as owner_disclosure
+from openexecutive.memory import episodic
+from openexecutive.orchestrator import disclosure
+from openexecutive.orchestrator.mcp_gateway import _with_disclosure
+from openexecutive.people import registry as people_registry
+from openexecutive.people import store as people_store
+
+HEADERS = {"x-caller-email": OWNER}
+
+
[email protected](autouse=True)
+def db(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> Iterator[Path]:
+ path = tmp_path / "episodic.db"
+ monkeypatch.setattr(episodic, "DB_PATH", path)
+ monkeypatch.setattr(people_store, "DB_PATH", path)
+ episodic.initialize_db(path)
+ people_store.initialize_db(path)
+ monkeypatch.setattr("openexecutive.audit.log_event", lambda *a, **k: None)
+ monkeypatch.setattr(disclosure, "_company_name", lambda: "Acme Corp")
+ for var in (
+ "AI_DISCLOSURE", "AI_DISCLOSURE_MAIL_TEXT", "AI_DISCLOSURE_CHAT_TEXT",
+ "AI_DISCLOSURE_HANDLE_IT", "AI_DISCLOSURE_APPROVED_SENDS",
+ ):
+ monkeypatch.delenv(var, raising=False)
+ monkeypatch.setenv("EXEC_DISPLAY_NAME", "Ava")
+ people_store.upsert_person(full_name="Olivia Owner", is_principal=True, email=OWNER)
+ people_registry.invalidate()
+ disclosure.reset()
+ yield path
+ people_registry.invalidate()
+ disclosure.reset()
+
+
+def _env(monkeypatch: pytest.MonkeyPatch, **values: str) -> None:
+ for key, value in values.items():
+ monkeypatch.setenv(key, value)
+
+
+# ── the Executive's mail ──────────────────────────────────────────────────────
+
+
+def test_the_mail_line_names_the_executive_the_company_and_the_owner() -> None:
+ assert disclosure.mail_line() == (
+ "Ava is an AI assistant working for Acme Corp. To reach a person, reply and ask, "
+ "or write to [email protected]."
+ )
+ body = disclosure.with_mail_line("Hi Joe,\n\nThanks.\n\nAva\n", recipients=["[email protected]"])
+ assert body == f"Hi Joe,\n\nThanks.\n\nAva\n\n---\n{disclosure.mail_line()}"
+ # Once: a body that already has it is left alone.
+ assert disclosure.with_mail_line(body, recipients=["[email protected]"]) == body
+
+
+def test_mail_only_to_the_owner_goes_without_it() -> None:
+ assert disclosure.with_mail_line("Your brief.", recipients=["[email protected]"]) == "Your brief."
+ assert disclosure.mail_line() in disclosure.with_mail_line("Hi", recipients=[OWNER, "[email protected]"])
+
+
+def test_html_mail_gets_an_escaped_paragraph() -> None:
+ out = disclosure.with_mail_line("<p>Hi</p>", recipients=["[email protected]"], as_html=True)
+ assert out.startswith("<p>Hi</p><hr><p")
+ assert "Acme Corp" in out and "<script" not in out
+
+
+def test_a_rewording_must_still_say_it_is_an_ai(monkeypatch: pytest.MonkeyPatch) -> None:
+ _env(monkeypatch, AI_DISCLOSURE_MAIL_TEXT="Sent by Ava for Acme.")
+ assert "AI assistant" in disclosure.mail_line()
+ _env(monkeypatch, AI_DISCLOSURE_MAIL_TEXT="Ava is Acme's AI helper.", AI_DISCLOSURE_CHAT_TEXT="Ava (AI)")
+ assert disclosure.mail_line() == "Ava is Acme's AI helper."
+ assert disclosure.chat_line() == "Ava (AI)"
+
+
+def test_off_nothing_is_added(monkeypatch: pytest.MonkeyPatch) -> None:
+ _env(monkeypatch, AI_DISCLOSURE="false")
+ assert disclosure.with_mail_line("Hi", recipients=["[email protected]"]) == "Hi"
+ assert disclosure.with_chat_line("Hi", "slack:s1") == "Hi"
+ assert _with_disclosure("google_workspace__send_gmail_message", "google_workspace__send_gmail_message",
+ {"to": "[email protected]", "body": "Hi"})["body"] == "Hi"
+
+
[email protected]("tool", ["google_workspace__send_gmail_message", "google_workspace__draft_gmail_message"])
+def test_gmail_sends_and_drafts_get_it(tool: str) -> None:
+ sent = {"to": "[email protected]", "subject": "Re: terms", "body": "Thanks."}
+ out = _with_disclosure(tool, tool, sent)
+ assert out["body"].endswith(disclosure.mail_line())
+ assert sent["body"] == "Thanks." # a copy: what the gates and records saw stays
+ html = _with_disclosure(tool, tool, {**sent, "body": "<p>Thanks.</p>", "body_format": "html"})
+ assert html["body"].startswith("<p>Thanks.</p><hr>")
+ to_owner = _with_disclosure(tool, tool, {**sent, "to": OWNER})
+ assert to_owner["body"] == "Thanks."
+
+
+def test_outlook_sends_replies_and_drafts_get_it() -> None:
+ message = {
+ "subject": "Plan",
+ "body": {"contentType": "Text", "content": "Attached."},
+ "toRecipients": [{"emailAddress": {"address": "[email protected]"}}],
+ }
+ sent = _with_disclosure("microsoft_365__send-mail", "microsoft_365__send_mail",
+ {"body": {"Message": message, "SaveToSentItems": True}})
+ assert sent["body"]["SaveToSentItems"] is True
+ assert sent["body"]["Message"]["body"]["content"].endswith(disclosure.mail_line())
+ reply = _with_disclosure("microsoft_365__reply-mail-message", "microsoft_365__reply_mail_message",
+ {"messageId": "m1", "body": {"Comment": "Sounds good."}})
+ assert reply["body"]["Comment"].startswith("Sounds good.<hr><p")
+ assert disclosure.mail_line() in reply["body"]["Comment"]
+ draft = _with_disclosure("microsoft_365__create-draft-email", "microsoft_365__create_draft_email",
+ {"body": message})
+ assert draft["body"]["body"]["content"].endswith(disclosure.mail_line())
+ # Sending a draft already made sends it as it is.
+ args = {"messageId": "d1"}
+ assert _with_disclosure("microsoft_365__send-draft-message", "microsoft_365__send_draft_message", args) is args
+ # A shape it doesn't know is left as it is.
+ odd: dict[str, Any] = {"body": "plain"}
+ assert _with_disclosure("microsoft_365__send-mail", "microsoft_365__send_mail", odd) is odd
+
+
+def test_html_that_could_hide_the_line_is_refused() -> None:
+ tool = "google_workspace__send_gmail_message"
+ for hiding in ("<p>Hi</p><!--", '<p>Hi</p><div style="display: none">', "<p>Hi</p><style>",
+ '<span style="font-size:0">'):
+ out = _with_disclosure(tool, tool, {"to": "[email protected]", "body": hiding, "body_format": "html"})
+ assert isinstance(out, str) and "plain text" in out
+ fine = _with_disclosure(tool, tool, {"to": "[email protected]", "body": '<p style="font-size:0.9em">Hi</p>',
+ "body_format": "html"})
+ assert isinstance(fine, dict)
+ # The owner's own HTML (a brief) goes as it is.
+ own = _with_disclosure(tool, tool, {"to": OWNER, "body": "<style>p{}</style>", "body_format": "html"})
+ assert isinstance(own, dict) and own["body"] == "<style>p{}</style>"
+ reply = _with_disclosure("microsoft_365__reply-mail-message", "microsoft_365__reply_mail_message",
+ {"messageId": "m1", "body": {"Comment": "Ok <!--"}})
+ assert isinstance(reply, str)
+
+
+def test_the_line_counts_only_at_the_end_of_what_shows() -> None:
+ line = disclosure.mail_line()
+ quoted = f"New reply.\n\nOn Monday Ava wrote:\n> Old reply.\n> ---\n> {line}\n"
+ assert disclosure.with_mail_line(quoted, recipients=["[email protected]"]).endswith(f"---\n{line}")
+ hidden = f"<p>New reply.</p><p hidden-title='{line}'></p>"
+ assert disclosure.with_mail_line(hidden, recipients=["[email protected]"], as_html=True).endswith("</p>")
+ assert disclosure.with_mail_line(hidden, recipients=["[email protected]"], as_html=True).count("<hr>") == 1
+
+
+def test_the_bare_line_names_nobody() -> None:
+ tool = "google_workspace__send_gmail_message"
+ out = _with_disclosure(tool, tool, {"to": "[email protected]", "body": "Thanks."}, bare=True)
+ assert isinstance(out, dict)
+ assert out["body"] == f"Thanks.\n\n---\n{disclosure.BARE_MAIL_LINE}"
+ assert "Acme" not in out["body"] and OWNER not in out["body"]
+
+
+# ── chat ──────────────────────────────────────────────────────────────────────
+
+
+def test_chat_gets_it_once_a_day_per_conversation() -> None:
+ first = disclosure.with_chat_line("Two exceptions were approved.", "slack:s1", now=1000.0)
+ assert first == "Two exceptions were approved.\n\n_Ava · AI assistant_"
+ assert disclosure.with_chat_line("Dana and the CFO.", "slack:s1", now=2000.0) == "Dana and the CFO."
+ assert disclosure.with_chat_line("Hello.", "slack:s2", now=2000.0).endswith("_Ava · AI assistant_")
+ later = 1000.0 + disclosure.CHAT_REPEAT + 1
+ assert disclosure.with_chat_line("Morning.", "slack:s1", now=later).endswith("_Ava · AI assistant_")
+ assert disclosure.with_chat_line("", "slack:s3") == ""
+
+
+def test_agent_messages_to_a_channel_carry_it(monkeypatch: pytest.MonkeyPatch) -> None:
+ import asyncio
+
+ from openexecutive.orchestrator import broadcast_tools
+
+ posted: list[str] = []
+
+ async def post(channel_id: str, text: str) -> dict[str, Any]:
+ posted.append(text)
+ return {"status": "sent"}
+
+ monkeypatch.setattr(broadcast_tools, "_post_slack_channel", post)
+ asyncio.run(broadcast_tools._dispatch("slack", "C1", "Q3 burn trending high"))
+ asyncio.run(broadcast_tools._dispatch("slack", "C1", "And again"))
+ assert posted == ["Q3 burn trending high\n\n_Ava · AI assistant_", "And again"]
+
+
+# ── replies sent as a person ──────────────────────────────────────────────────
+
+
+def test_each_person_starts_from_the_install_and_can_change_theirs(monkeypatch: pytest.MonkeyPatch) -> None:
+ start = owner_disclosure.get(7)
+ assert start.handle_it is True and start.approved_sends is False
+ owner_disclosure.set_(7, approved_sends=True, updated_by="t")
+ assert owner_disclosure.get(7).approved_sends is True
+ owner_disclosure.set_(8, handle_it=False, updated_by="t")
+ _env(monkeypatch, AI_DISCLOSURE_HANDLE_IT="false", AI_DISCLOSURE_APPROVED_SENDS="true")
+ assert owner_disclosure.get(9) == owner_disclosure.Disclosure(9, handle_it=False, approved_sends=True)
+ assert owner_disclosure.get(8).handle_it is False # their own choice stays theirs
+ # A switch they never named still follows the install.
+ assert owner_disclosure.get(8).approved_sends is True
+ _env(monkeypatch, AI_DISCLOSURE_HANDLE_IT="true")
+ assert owner_disclosure.get(7).handle_it is True
+
+
+def test_sign_adds_the_line_by_the_kind_of_send(monkeypatch: pytest.MonkeyPatch) -> None:
+ line = owner_disclosure.LINE
+ assert owner_disclosure.sign(7, "Thanks,\nOlivia", on_its_own=True) == f"Thanks,\nOlivia\n\n{line}"
+ assert owner_disclosure.sign(7, "Thanks,\nOlivia", on_its_own=False) == "Thanks,\nOlivia"
+ owner_disclosure.set_(7, approved_sends=True, updated_by="t")
+ assert owner_disclosure.sign(7, "Thanks", on_its_own=False).endswith(line)
+ assert owner_disclosure.sign(7, f"Thanks\n\n{line}", on_its_own=False).count(line) == 1
+ _env(monkeypatch, AI_DISCLOSURE="false")
+ assert owner_disclosure.sign(7, "Thanks", on_its_own=True) == "Thanks"
+
+
+def client(monkeypatch: pytest.MonkeyPatch) -> TestClient:
+ async def status(email: str | None, *, gmail: Any = None) -> str:
+ return "connected"
+
+ monkeypatch.setattr(route, "gmail_status", status)
+ monkeypatch.setenv("OE_LOCAL_LOGIN", "1")
+ monkeypatch.delenv("OE_PUBLIC_DEPLOYMENT", raising=False)
+ app = FastAPI()
+ app.include_router(route.router)
+ return TestClient(app)
+
+
+def test_the_route_shows_and_changes_the_callers_own_switches(client: TestClient) -> None:
+ body = client.get("/delegation", headers=HEADERS).json()
+ assert body["disclosure"] == {
+ "enabled": True, "handle_it": True, "approved_sends": False, "line": owner_disclosure.LINE,
+ }
+ body = client.put("/delegation/disclosure", json={"handle_it": False}, headers=HEADERS).json()
+ assert body["disclosure"]["handle_it"] is False
+ principal = people_store.find_principal_person()
+ assert principal is not None and principal.id is not None
+ assert owner_disclosure.get(principal.id).handle_it is False
+ assert client.put("/delegation/disclosure", json={"other": True}, headers=HEADERS).status_code == 422
+
+
+def test_turning_one_off_needs_a_caller_the_api_can_vouch_for(
+ client: TestClient, monkeypatch: pytest.MonkeyPatch
+) -> None:
+ monkeypatch.delenv("OE_LOCAL_LOGIN", raising=False)
+ resp = client.put("/delegation/disclosure", json={"handle_it": False}, headers=HEADERS)
+ assert resp.status_code == 409 and resp.json()["detail"]["code"] == "caller_signing_required"
+ resp = client.put("/delegation/disclosure", json={"approved_sends": True}, headers=HEADERS)
+ assert resp.status_code == 200 and resp.json()["disclosure"]["approved_sends"] is True
diff --git a/packages/core/tests/unit/test_bot_link_previews.py b/packages/core/tests/unit/test_bot_link_previews.py
index f1710125..50643c61 100644
--- a/packages/core/tests/unit/test_bot_link_previews.py
+++ b/packages/core/tests/unit/test_bot_link_previews.py
@@ -35,7 +35,7 @@ def test_telegram_messages_turn_previews_off(monkeypatch: pytest.MonkeyPatch) ->
def test_slack_replies_do_not_unfurl() -> None:
source = inspect.getsource(slack_bot)
- assert "await say(text=response, thread_ts=thread_ts, unfurl_links=False, unfurl_media=False)" in source
+ assert "thread_ts=thread_ts, unfurl_links=False, unfurl_media=False," in source
def test_discord_replies_suppress_embeds() -> None:
diff --git a/packages/core/tests/unit/test_delegation_handle_it.py b/packages/core/tests/unit/test_delegation_handle_it.py
index 72a6dd5a..5848d099 100644
--- a/packages/core/tests/unit/test_delegation_handle_it.py
+++ b/packages/core/tests/unit/test_delegation_handle_it.py
@@ -129,6 +129,34 @@ def test_off_it_leaves_a_card_as_before(owner: Any, models: dict[str, Any]) -> N
# ── sending on its own ────────────────────────────────────────────────────────
+def test_a_reply_sent_on_its_own_says_an_ai_wrote_it(
+ owner: Any, models: dict[str, Any], monkeypatch: pytest.MonkeyPatch
+) -> None:
+ from openexecutive.delegation import disclosure
+
+ monkeypatch.delenv("AI_DISCLOSURE", raising=False)
+ _on(owner)
+ mailbox, result = _scan_one(owner)
+ assert result.handled == 1
+ (spec,) = mailbox.specs
+ assert spec.body.endswith(disclosure.LINE)
+ assert inbox.card_payload(_only_card(owner))["draft_body"] == spec.body
+
+
+def test_a_reply_waiting_for_them_has_no_ai_line_unless_they_asked(
+ owner: Any, models: dict[str, Any], monkeypatch: pytest.MonkeyPatch
+) -> None:
+ from openexecutive.delegation import disclosure
+
+ monkeypatch.delenv("AI_DISCLOSURE", raising=False)
+ mailbox, _ = _scan_one(owner)
+ assert disclosure.LINE not in mailbox.specs[0].body
+ disclosure.set_(owner.id, approved_sends=True, updated_by="test")
+ mailbox, _ = _scan_one(owner, _msg("m2", "t2"))
+ assert mailbox.specs[0].body.endswith(disclosure.LINE)
+
+
+
def test_on_it_sends_a_reply_to_someone_they_know(db: Path, owner: Any, models: dict[str, Any]) -> None:
_on(owner)
mailbox, result = _scan_one(owner)
diff --git a/packages/core/tests/unit/test_roster_intake.py b/packages/core/tests/unit/test_roster_intake.py
index e10ee855..0209d170 100644
--- a/packages/core/tests/unit/test_roster_intake.py
+++ b/packages/core/tests/unit/test_roster_intake.py
@@ -98,6 +98,18 @@ def test_the_granted_acknowledgement_reaches_an_address_off_the_roster(roster: S
assert args["to"] == STRANGER and args["body"] == roster_intake.ACK_TEXT
+def test_the_acknowledgement_says_it_is_an_ai_without_naming_anyone(
+ roster: SimpleNamespace, monkeypatch: pytest.MonkeyPatch
+) -> None:
+ from openexecutive.orchestrator import disclosure
+
+ monkeypatch.delenv("AI_DISCLOSURE", raising=False)
+ gateway, sent = _gateway()
+ asyncio.run(roster_intake.send_email_ack(gateway, STRANGER))
+ body = sent.await_args.args[1]["arguments"]["body"]
+ assert body == f"{roster_intake.ACK_TEXT}\n\n---\n{disclosure.BARE_MAIL_LINE}"
+
+
@pytest.mark.parametrize("over", [
{"body": "Anything else at all"},
{"subject": "Re: your invoice"},
diff --git a/packages/ui/src/app/delegate/act-as-me/page.tsx b/packages/ui/src/app/delegate/act-as-me/page.tsx
index dd1de8c6..ce2026e0 100644
--- a/packages/ui/src/app/delegate/act-as-me/page.tsx
+++ b/packages/ui/src/app/delegate/act-as-me/page.tsx
@@ -3,6 +3,7 @@
import ActAsMeCard, { ACT_AS_ME_INTRO } from "@/components/settings/ActAsMeCard";
import { useDelegation } from "@/components/settings/HandleItCard";
import DelegatePage from "@/components/delegate/DelegatePage";
+import ReplyDisclosureCard from "@/components/settings/ReplyDisclosureCard";
import { LearnedCard, SuggestedActionsCard, TrainingNote } from "@/components/settings/TrainingCard";
// Delegate → Act as me: drafts written as you, in your own mailbox, and how
@@ -13,12 +14,14 @@ import { LearnedCard, SuggestedActionsCard, TrainingNote } from "@/components/se
export default function ActAsMePage() {
const load = useDelegation();
const training = load.state === "ready" ? load.settings?.training : null;
+ const disclosure = load.state === "ready" ? load.settings?.disclosure : null;
return (
<DelegatePage feature="act_as_me" description={ACT_AS_ME_INTRO}>
{training && <TrainingNote />}
<ActAsMeCard onSettings={load.setSettings}>
{training && <SuggestedActionsCard training={training} onSettings={load.setSettings} />}
{training && <LearnedCard training={training} onSettings={load.setSettings} />}
+ {disclosure && <ReplyDisclosureCard disclosure={disclosure} onSettings={load.setSettings} />}
</ActAsMeCard>
</DelegatePage>
);
diff --git a/packages/ui/src/components/settings/ReplyDisclosureCard.tsx b/packages/ui/src/components/settings/ReplyDisclosureCard.tsx
new file mode 100644
index 00000000..90de0c6e
--- /dev/null
+++ b/packages/ui/src/components/settings/ReplyDisclosureCard.tsx
@@ -0,0 +1,114 @@
+"use client";
+
+import { useState } from "react";
+
+import SettingsCard from "@/components/settings/SettingsCard";
+import { setReplyDisclosure, type DelegationSettings, type ReplyDisclosure } from "@/lib/api";
+
+type Switch = "handle_it" | "approved_sends";
+
+const SWITCHES: { key: Switch; label: string; help: string }[] = [
+ {
+ key: "handle_it",
+ label: "Sent on their own (Handle it for me)",
+ help: "No one reads these before they go.",
+ },
+ {
+ key: "approved_sends",
+ label: "Drafts you approve and send",
+ help: "You read these first, so they go out as you wrote them unless you turn this on.",
+ },
+];
+
+// On Delegate → Act as me: your own switches for the AI line on replies sent
+// as you (PUT /delegation/disclosure, delegation/disclosure.py). The line is
+// written into each reply when it's drafted, so a change applies to the next
+// one. Turning off the Handle it line asks first. Nothing shows when the
+// install has AI disclosure off.
+export default function ReplyDisclosureCard({
+ disclosure,
+ onSettings,
+}: {
+ disclosure: ReplyDisclosure;
+ onSettings: (next: DelegationSettings) => void;
+}) {
+ const [busy, setBusy] = useState(false);
+ const [error, setError] = useState<string | null>(null);
+ const [confirmOff, setConfirmOff] = useState(false);
+
+ if (!disclosure.enabled) return null;
+
+ const save = async (key: Switch, value: boolean) => {
+ setBusy(true);
+ setError(null);
+ try {
+ onSettings(await setReplyDisclosure({ [key]: value }));
+ setConfirmOff(false);
+ } catch (err) {
+ setError(err instanceof Error ? err.message : "Could not save the setting.");
+ } finally {
+ setBusy(false);
+ }
+ };
+
+ const toggle = (key: Switch, value: boolean) => {
+ if (key === "handle_it" && !value) {
+ setConfirmOff(true);
+ return;
+ }
+ void save(key, value);
+ };
+
+ return (
+ <SettingsCard
+ title="Replies sent from your mailbox"
+ description={`Choose which replies end with “${disclosure.line}”, so people can tell your AI assistant wrote them.`}
+ >
+ <div className="flex flex-col gap-4">
+ {SWITCHES.map((s) => (
+ <label key={s.key} className="flex cursor-pointer items-start gap-3">
+ <input
+ id={`reply-disclosure-${s.key}`}
+ type="checkbox"
+ checked={disclosure[s.key]}
+ disabled={busy}
+ onChange={(e) => toggle(s.key, e.target.checked)}
+ className="mt-0.5 h-5 w-5 flex-shrink-0 accent-indigo-500"
+ />
+ <span>
+ <span className="block text-[15px] font-semibold text-fg">{s.label}</span>
+ <span className="mt-0.5 block text-sm text-fg-muted leading-snug">{s.help}</span>
+ </span>
+ </label>
+ ))}
+ {confirmOff && (
+ <div className="rounded-md border border-line p-3 text-sm">
+ <p className="text-fg">
+ Replies Handle it for me sends on their own will no longer say your AI assistant wrote them.
+ </p>
+ <div className="mt-3 flex flex-wrap gap-2">
+ <button
+ type="button"
+ disabled={busy}
+ onClick={() => void save("handle_it", false)}
+ className="rounded-md border border-line px-3 py-1.5 font-medium text-fg"
+ >
+ Turn it off
+ </button>
+ <button
+ type="button"
+ disabled={busy}
+ onClick={() => setConfirmOff(false)}
+ className="rounded-md px-3 py-1.5 text-fg-muted"
+ >
+ Keep it on
+ </button>
+ </div>
+ </div>
+ )}
+ <p className="text-sm text-fg-muted">A change applies to the next reply it writes.</p>
+ </div>
+ {error && <p className="mt-2 text-sm text-red-500">{error}</p>}
+ </SettingsCard>
+ );
+}
diff --git a/packages/ui/src/lib/api.ts b/packages/ui/src/lib/api.ts
index 7e21d29a..11c95208 100644
--- a/packages/ui/src/lib/api.ts
+++ b/packages/ui/src/lib/api.ts
@@ -1534,6 +1534,8 @@ export interface DelegationSettings {
handle_it?: HandleIt | null;
// Absent (or null) on a backend that predates per-setting training.
training?: Training | null;
+ // Absent (or null) on a backend that predates the AI line on replies.
+ disclosure?: ReplyDisclosure | null;
// The owner's "Let team members use Act as me", while the install allows
// it; null (or absent) for everyone else.
team?: DelegationTeam | null;
@@ -1602,6 +1604,16 @@ export interface Training {
learned: TrainingLearned[];
}
+// The AI line on replies sent as you (delegation/disclosure.py): on replies
+// Handle it for me sends on its own, and on drafts you send. `enabled` is the
+// install's AI_DISCLOSURE; when it's off neither line is added.
+export interface ReplyDisclosure {
+ enabled: boolean;
+ handle_it: boolean;
+ approved_sends: boolean;
+ line: string;
+}
+
export interface HandleIt {
enabled: boolean;
mode: HandleItMode;
@@ -1855,6 +1867,19 @@ export async function setTraining(update: Partial<Record<TrainingSetting, boolea
return (await res.json()) as DelegationSettings;
}
+export async function setReplyDisclosure(update: {
+ handle_it?: boolean;
+ approved_sends?: boolean;
+}): Promise<DelegationSettings> {
+ const res = await fetch(`${API_BASE}/delegation/disclosure`, {
+ method: "PUT",
+ headers: { "Content-Type": "application/json" },
+ body: JSON.stringify(update),
+ });
+ if (!res.ok) throw await delegationError(res, "Couldn't change the AI line on your replies.");
+ return res.json();
+}
+
export async function setHandleIt(update: {
enabled?: boolean;
mode?: HandleItMode;