Skip to content

Build AI methods with your coding agent, then run them anywhere: as an MCP for chatbots, as a webapp for people, or via API for your software.

Drift Hunt: campaign close-out — records, handoff, and tracker (#1046)

Louis Choquelbdefdaa59b37
diff --git a/TODOS.md b/TODOS.md
index 46d252a54..205d87ff9 100644
--- a/TODOS.md
+++ b/TODOS.md
@@ -1,241 +1,173 @@
-# Subject Grants — dropping the generic positional-subject exception
+# Drift Hunt — campaign tracker
-Hardens the keyword-only-arguments convention: Exception 1 (the positional subject) stops being a generic permission and becomes an **explicitly granted, recorded exception**. There is no separate design doc — this file is the authority on the design (decisions, rubric, registry format) and tracks execution state.
+A campaign, not a feature: sweep every hand-written doc page for code↔doc drift, fix what's stale, and come out with per-section defect-density evidence. The drift-contracts system (`docs/contribute/drift-contracts.md`, design at `wip/drift-contracts-design.md`) enforces the floor *going forward*; this hunt backfills the floor once. Promoted from `wip/kick-off-drift-hunt.md` (now a pointer stub) — this file is the live tracker.
## Cold-start context (update at every checkpoint)
-- **Status:** **TRACK COMPLETE (2026-07-11). Phase 5 finalization DONE with Louis' greenlight:** `--seed` + `seeded` schema support deleted (registry schema strict: exactly `param` + `rationale`), CHANGELOG `[Unreleased]` carries the convention entry + the breaking demoted-signatures note, convention doc post-transitional, all three open drift contracts reviewed + acked (cli-docs, keyword-only-convention, config-docs), demoted-public-surfaces inventory handed off to `wip/subject-grants/demoted-public-surfaces.md` for the release-wave cross-repo sweep. Registry: 1,742 grants, no seeded field. Phase M history: regrind waves A/B/C + checkpoint review triaged (see the wip record). **PR #1040 → dev OPEN.** The regrind ran as three waves on top of the reseed commit `d58cd847b`: **Wave A `fb1a91197`** — AST classifier compared each seeded def's signature across pre-grind base / archive tip / merged tree; 1,470 archive-KEPT rationales ported verbatim. **Wave B `65a4153ae`** — 148 archive demotes re-applied by splicing the archive's demoted headers byte-for-byte, PLUS the @override/protocol implementers the archive demoted alongside (registry-invisible carve-outs), recovered pyright-guided; call sites re-labelled mechanically; full agent-test green. **Wave C `f83308367`** — genuine review of the dev-delta (285 new + 24 dev-changed): 268 kept, 41 demoted per settled case-law families; GraphTracerProtocol.setup demoted with parity (GraphTracer + GraphTracerNoOp); full agent-test green. Method + judgment calls recorded in `wip/subject-grants/regrind-after-dev-merge.md`. Prior merge context: rewind to pre-grind `9f5bd54a3` + merge `5bb8c2756` (dev tip `048c87971`) + reseed `d58cd847b`; full first grind archived at tag **`subject-grants-grind-archive`** (`2ebf3f293` — checkpoint-2/3 records live in that branch's wip/). **PUSH STATE: NOT pushed; rewind ⇒ force-push (Louis-approved), push the archive tag too.**
-- **Resume point:** nothing left on this branch — the track is complete and pushed to PR #1040. Remaining work is release-gated and lives outside this branch: the cross-repo sweep over `wip/subject-grants/demoted-public-surfaces.md` once the release ships.
-- **Registry state (post-regrind + checkpoint triage):** **1,743 grants total, 0 seeded.** Wave B deleted 148 entries, wave C deleted 41, checkpoint triage deleted 3 more; all remaining rationales are real (ported or newly written). The classifier + rewriter scripts live in the session scratchpad (throwaway; method documented in the wip record).
-- **Rubric case-law built up over batches 1–7 (apply consistently to cli/cogt/plugins/tools):**
- - DEMOTE patterns: key+payload attachment (`set_event_log(context_key, *, event_log)` — the payload is the verb's object, the positional is a registry/target key); field-bag factories (`ConceptFactory.make(concept_code, *, domain_code, description, …)` — first field arbitrarily positional); mode/option params (`to_dict(disclosure_mode)`, `render_stuff_spec(output_format)`, `rendered_for_prompt(text_format)`, `pretty_print_*(title)`); instrumental/context params (`_evaluate_expression(working_memory)`, `_start_pipe_span(parent_otel_context)`, `render_with_images(registry)`, `_emit_via_registered_context(context)`); optional-selector-among-alternatives (`build_inputs_for_pipe(pipe_code=None, *, mthds_contents, bundle_path)`); symmetric/near-symmetric pairs (`expected`/`actual`, `domain_path`+`local_code`, `trace_name`+`trace_name_redacted`, and **same-type operand pairs** like `is_compatible(tested_concept: Concept, *, wanted_concept: Concept)` — two `Concept`s, no single candidate, demote like `copy_file(source, target)`); lone recursion accumulators (`needed_inputs(visited_pipes)`); config-first (`pipeline_run_setup(execution_config)`) and dependency-first (`load_telemetry_config(secrets_provider)`); filter/scope params (`list_models(categories)`, `validate_all(library_dirs)`); bare-literal call sites (`_get_config_file_not_found_error_msg("routing profile library")`); derived-value lookups (`_get_structure_class_import(class_name)`).
- - KEEP families: entity-keyed getters/setters/managers (hub, libraries, registries, working memory, env — `get_stuff(name)`, `set_config(config)`, `open_tracer(graph_id)`); predicates (`is_x(value)`); single-operand transformations/conversions/parsers/validators/normalizers; factories-from-source (`make_from_blueprint(blueprint)`); protocol event/payload handlers (`observe_before_run(payload)`, `run(pipe_job)`, `emit(event)`, `on_pipe_end_success(node_id, *, …)`); positional-Callable protocols (visitors, re.sub repl, tenacity before_sleep, threading.excepthook, decorators, sort keys); pytest hooks (pluggy binds by name); noun-named single-operand derivations (`page_slug(cls)`, `_make_pk(pipeline_run_id)` — id encoded, not looked up); type-directed accessors (`main_stuff_as(content_type)`, `get_items(item_type)`).
- - **job_metadata run-family demoted hierarchy-wide** (JobMetadata = information, not operand): PipeAbstract run/validate/span family + all operator/controller/signature overrides + PipeFactoryProtocol.make + factory impls. ⚠ cogt batch: `dry_mock.py`'s `__call__(self, job_metadata, *, …)` reporter protocol was deliberately LEFT positional — decide that family (report_dry_llm_job / report_mock_usage_llm_job must match the protocol) in the cogt batch; `update_job_metadata(job_metadata)` is a genuine keep (metadata IS the operand there).
- - Registry edits are done with a tomlkit script per batch (`load_toml_with_tomlkit(path)` / `save_toml_to_path(data, path=…)`); keeps = replace rationale + drop `seeded`, demotes = delete entry, guard's `dead-grant` full-scan confirms nothing dangles. Rationale style: terse but def-specific family formulas (explicitly allowed by the rubric). (Single-entry deletions can be a plain line-block delete — deletion preserves sort order; agent-check's plxt-format + cko validate the result.)
- - **Batches 8–11 learnings (plugins/cli/cogt/tools — now settled case-law):** (a) **console/logger/registry/client/secrets_provider as first param = instrumental sink/dependency → DEMOTE** even when it's the sole param — the object is in the verb name (`display_gateway_accepted_message(console)`, `InferenceBackendLibrary.load(secrets_provider)` matches the demoted `load_telemetry_config(secrets_provider)`). Consistency check: no reviewed package kept a console-as-sink grant (`PipelexHub.set_console(console)` is the lone exception — there console IS the operand being installed). (b) **`X | None = None` first positional that is a scope/location/destination/override and is often omitted or keyword at call sites → DEMOTE** (`check_config_files(config_dir=None)`, `make_pipelex_for_agent_cli(library_dirs=None)`, `generate_error_pages_cmd(output=None)`). But an **optional param that is genuinely the operand being resolved/parsed/acted-upon stays KEEP** (`_resolve_repo_root(repo_root)`, `_parse_config_arg(config_arg)`, `_end_otel_span_with_error(span)` — span is the object ended; `drift_plan_cmd(contract_id)` — contract_id is the direct object). Optionality is a *signal*, not a verdict — the discriminator is the verb-object/single-candidate test. (c) **optional-selector-among-alternatives → DEMOTE** (`make_prompt_document(uri=None, *, base64_data, raw_bytes)` — three co-equal sources; matches demoted `build_inputs_for_pipe`). (d) **positional-Callable protocols → KEEP** and this bit hard in cogt: `report_dry_llm_job`/`report_mock_usage_llm_job` are assigned as a `report_func` and called `report_func(job_metadata, ...)` positionally to satisfy `_ReportLLMJobFunc.__call__(job_metadata, *, ...)` — so job_metadata STAYS positional there (the hierarchy-wide job_metadata demote does NOT apply to protocol-satisfying reporters). Same KEEP for jinja `finalize`, `re.sub` repl, tenacity `should_retry`/`_transport_retry_wait`, Rich `__rich_repr__`, decorators (`update_job_metadata(func)`). (e) **mode/strategy selector as sole param → DEMOTE** (`get_aws_access_keys_with_method(api_key_method)` — enum match/case). (f) **bare-literal label param → DEMOTE** (`_validate_version("mthds-ui", value=...)` — name is only for the error msg, value is the real subject). (g) **Mechanical tooling that worked well (reuse for Phase-5-style sweeps):** a signature rewriter that moves the first non-self/cls positional after `*` (AST-reconstructs the param list preserving defaults/annotations; handles multiline) + a call-site fixer that inserts `<param>=` before the first positional arg at every call keyed by simple function name, run over `pipelex/` AND `tests/`, then pyright drives the checklist of remaining positional call sites. Scripts in the session scratchpad; `fko` handles single-param no-`*` defs, the rewriter handles `subject, *, opt` shapes fko can't.
- - **Checkpoint-2 review learnings (apply to cli/cogt/plugins/tools):** (a) **Protocol/ABC parity is mandatory when demoting** — if a def implements a `@runtime_checkable` Protocol or overrides an ABC, demote the interface AND every implementer together (mode-param demotes on `StuffContent`/`StuffArtefact.rendered_for_template_async` left the `TextFormatRenderable` protocol positional → type-checker-blind landmine; caught at checkpoint 2). When a demote touches a class in a not-yet-reviewed batch's file (its protocol/base), reach in and demote it too (the ImageRenderable spillover precedent). (b) `rendered_for_template_async(text_format)` joins the mode/option DEMOTE family alongside `rendered_for_prompt(text_format)`. (c) **cli batch action item:** `cli/agent_cli/commands/pipe_cmd.py::_add_type_specific_fields` (seeded) is a near-duplicate of the already-demoted `builder/operations/pipe_ops.py::add_type_specific_fields` — demote the cli twin to match. (d) A grant whose call sites ALL pass by keyword isn't automatically a demote (grants permit but don't require positional) — the disqualifier is the *single-candidate* test, not call-site style.
-- **Batch execution recipe (how to grind one package — cli/cogt/plugins/tools):**
- 1. `.venv/bin/pipelex-dev check-keyword-only --report` — confirm the package's seeded count (and that violations = 0 before you start).
- 2. **Enumerate the batch's seeded entries.** Each `["pipelex/<pkg>/….py::<qualname>"]` with `seeded = true` in `subject_grants.toml` is one def to review. A python read is cleanest: `tomllib.load` the registry, filter keys `startswith("pipelex/<pkg>/")` where the entry still has `seeded`. (grep works too: entries are 3–4 lines, key → `param` → `rationale` → `seeded = true`.)
- 3. **Review each entry** against the Rubric + the case-law above. Open the def (`path::qualname` → file+symbol), read its signature, `git grep` its call sites, then pick exactly one: **KEEP** (replace the SEEDED placeholder with a real def-specific rationale, drop the `seeded` line) · **DEMOTE** (move the subject param after the `*` — insert one if absent — on the def AND every Protocol/ABC/override sibling together, fix call sites pyright-guided, delete the entry) · **HATCH** (rare: `# kw-only: ignore` on the def line for framework-positional shapes, delete the entry).
- 4. **Apply registry edits** with a tomlkit script (`from pipelex.tools.misc.toml_utils import load_toml_with_tomlkit, save_toml_to_path`) — batch all keeps/demotes for the package in one pass; keeps mutate `rationale` + delete the `seeded` key, demotes `del data[key]`. Single-entry deletions may instead be a plain line-block Edit (deletion preserves sort order). Never hand-roll tomlkit (D10).
- 5. **Gate:** `make agent-check` (runs fko → format → lint → pyright → mypy → cko; cko's dead-grant full-scan confirms nothing dangles). Run targeted tests for touched areas; run **full `make agent-test` if the batch demoted ANY framework-adjacent def** (Protocol/callback/Jinja/Typer/pytest-hook — pyright is blind to those, per the Risks section).
- 6. **Commit** one per batch: `Subject grants batch N: review <pkg> (X kept, Y demoted)`. Do NOT push. Update this file's Status line (seeded-remaining + commit SHA) as you go, or at the next checkpoint.
- ⚠ **Protocol/ABC parity (checkpoint-2 learning, bites silently):** demoting a def that implements a `@runtime_checkable` Protocol or overrides an ABC MUST demote the interface + every implementer in the same commit — even when the interface lives in a not-yet-reviewed file (reach in, `*`-it, delete its seeded grant). Leaving the interface positional while impls are keyword-only is a type-checker-blind landmine (exactly the two defects checkpoint 2 caught).
-- **Deviations from the plan (all deliberate):**
- - The seed NEVER enters literal-typed subjects (rather than seeding then removing them) — they can never be granted (D4), so seeding them would only create entries destined for deletion. Same end state, no dead entries.
- - `--report` grant/seeded progress meter lives in `check-keyword-only --report`; violation kinds are `missing-star` / `ungranted-subject` / `literal-subject` / `grant-param-mismatch` (def-level, lean path sees it) / `dead-grant` (full-scan only — covers deleted, renamed, demoted-to-all-keyword, and newly-carved-out defs).
- - `SubjectGrantRegistryError` is defined IN `keyword_only_guard.py` (not an `exceptions.py`): the lean hook path cannot import a sibling module without dragging the `pipelex` package chain. It is a plain `Exception` subclass, not a `PipelexError` (no error page needed; `make gep` not required).
- - TWO drift contracts opened (not one): `cli-docs` also triggered because the literal sweep touched `pipelex/cli/**`. Both acked genuinely; dogfood log has a real-catch (keyword-only-convention) + a clean-pass (cli-docs) entry.
- - The workspace-root `.claude/rules/python-standards.md` keyword-only section was spliced from the kit source by hand (section-only replacement — the rest of that file tracks dev, which is ahead of this branch on Python-3.10-drop wording; do NOT copy the whole kit file over it).
-- **Branch / worktree:** `feature/Signatures` in the `_sig` worktree (branched off `docs/Update`). Treat `_sig` as the repo root. This file replaced the drift-contracts tracker inherited from `docs/Update` — that plan lives on in the `docs/Update` branch history and is in its dogfood phase; not this project's concern.
-- **Problem being solved:** the guard mechanically allows ANY first non-`self`/`cls` parameter to stay positional-or-keyword. Coding agents abuse that judgment-shaped permission, leaving first arguments positional when they are not the subject or not obviously so. The check must stay deterministic, but the exception must become deliberate, recorded, and reviewable — and **every existing positional subject is considered new: seeded into the registry unreviewed, then actually checked** (Phase 4).
-- **The shape in one paragraph:** a committed registry (`subject_grants.toml` at repo root) lists every def allowed a positional subject, with the subject param name and a one-sentence rationale. `check-keyword-only` fails on any positional subject without a matching grant, on any grant without a matching def (staleness is symmetric), and on any literal-typed positional subject regardless of grant. `make fko` already rewrites to the all-keyword form, so the lazy path is the strict path; keeping a positional subject costs a deliberate `subject-grant` command with an honest rationale, visible as a registry diff in every PR. Judgment happens at grant time and in review of the registry diff — CI stays 100% deterministic (no LLM in any gate, ever).
-
-### Decisions (settled — do not relitigate without Louis)
-
-- **D1 — Generic exception dropped.** A positional-or-keyword subject is legal only when a subject grant exists for that def. No grant → violation; `fko` auto-fixes it to all-keyword.
-- **D2 — Registry** = `subject_grants.toml` at repo root. One entry per def, keyed `<relative_path>::<qualified_name>` (exactly the guard's `Violation.key` format), recording `param` (the subject's name) and `rationale`. The file is rewritten sorted by key on every change so diffs are stable and merge conflicts stay trivial.
-- **D3 — CI stays deterministic.** No model verdict in any gate. The guard only verifies grant existence, param match, and freshness. The judgment lives in the grant rationale and in PR review of the registry diff (drift-ack spirit: an honest sentence, on the record).
-- **D4 — Literal-typed subjects banned outright**, grant impossible: a subject annotated `bool`, `int`, or `float` (including `X | None` / `Optional[X]` forms of those) is a violation no matter what. `f(True)` call sites are never acceptable. `str` subjects stay grantable — they are the house style (`pipe_code`, `name`, `uri`, …).
-- **D5 — Strict-all scope.** Grants are required for every positional subject: lone-subject defs (`def render(node)`) and subject-plus-kwonly defs (`def f(spec, *, ...)`) alike. Relaxing later (e.g. exempting lone-subject defs) is trivial; re-tightening later reopens ~1,000 defs — so start strict.
-- **D6 — All existing positional subjects are treated as NEW.** Seeded into the registry with `seeded = true` and a placeholder rationale, then every one is genuinely reviewed in Phase 4: real rationale (keep), or converted to keyword-only (demote), or `# kw-only: ignore` (rare, framework-positional). Zero `seeded` entries may remain at finalization.
-- **D7 — Staleness is symmetric and hard-fails.** Positional subject without grant = violation. Grant whose def no longer exists, or whose recorded `param` no longer matches the def's first param = violation. Renames and file moves therefore force a re-grant — that is a deliberate re-decision, not friction to engineer away.
-- **D8 — Existing mechanisms unchanged.** `# kw-only: ignore` (stronger hatch, short-circuits before the rule — required for framework-positional callables), the symmetric-tuple allowlist (whole-function, stays curated in code, NOT merged into the registry — different semantics), all carve-outs, and `fko`'s insert-leftmost behavior. Note `fko`'s reach grows: an ungranted subject def is now a violation, so `fko` (which runs inside `make agent-check`) will silently kwonly it — grant BEFORE running checks if you want the subject positional.
-- **D9 — `/` (positional-only) stays banned.** The convention doc's deliberate no-`/` decision stands; callers keep keyword discretion on granted subjects.
-- **D10 — Hook budget preserved.** The lean single-file path (`keyword_only_guard.py` run by file path from the PostToolUse hook) stays stdlib-only: it reads the registry with `tomllib` (read-only, machine-written file — the tomli-error-attrs concern applies to user-config parsing, not here). Registry writes happen only in the full CLI via `pipelex/tools/misc/toml_utils.py` (`load_toml_with_tomlkit` / `save_toml_to_path` — do not hand-roll tomlkit).
-- **D11 — Same-qualname defs share one entry.** `@overload` stubs and conditional redefinitions collapse onto one key; the recorded `param` must match each of them (forces overload stubs to align their subject name — acceptable).
+- **Status:** **✅ CAMPAIGN DONE — CHECKPOINT 2 CLOSE-OUT EXECUTED (2026-07-12), STOPPED for Louis' final review.** The record: `wip/drift-hunt/findings/SUMMARY.md` (close-out section: coverage disclosure, Stage-2 resolution table, ranked drift-contract shortlist for the Phase 3 verdict, lessons) + `inventory.md`; the handoff: `deferred-code-findings.md` (10 items). Full `make agent-test` GREEN on the reconciled branch. `wip/drift-hunt/` pruned (removed files at `git show drift-hunt-archive:wip/drift-hunt/<path>` — pushed tag pinning `9a6201a5e`). ⚠ Record correction: the observer-provider-injection.md "partially fixed" note below was a **stale-base artifact** — the page is fully fixed on dev (S2-3/#1042 fixed Part-2 F38–F42; re-verified against live code at close-out), no follow-up needed. Everything below this point is historical narrative. — Historical: **ALL FOUR STAGE-2 PRs MERGED TO DEV (2026-07-12)** — squash SHAs: **#1042 (S2-3) `9814e2f4c`, #1043 (S2-1) `126616a88`, #1044 (S2-2) `46511ce9c`, #1045 (S2-4) `cb13e5b33`**. dev tip = `cb13e5b33`. All four PR branches deleted, remote AND local (local tips verified ancestors of the merged `refs/pull/N/head` before force-delete — squash merges always fail `git branch -d`'s ancestry check, that refusal is expected, not a danger signal). Merges were forced sequential by branch protection (head must be up to date with dev): after each squash the next PR went BEHIND → `gh pr update-branch` + squash auto-merge + wait for CI, three times. The main checkout `../pipelex` is back on `dev` at the tip; this `_drift` worktree stays on `docs/Drift-hunt`. Pre-merge, PR #1045 was finalized: Louis's review edits + bot-triage fixes landed as PR-branch commit `1b776eca5` (configure-ai-providers.md local-overrides/init/telemetry.toml notes; `generate-mthds-schema` added to the kit-rules no-make fallback list — verified against `Makefile:882` `plxt-lint: env generate-mthds-schema-quiet`, regenerated via `make rules`; data-extraction.md explicit `storage_dir` for pre-boot `LocalObserver` — verified against `local_observer.py:19`; observer-provider-injection.md JSONL filenames corrected to the `LocalObserverEventType` values); all 8 bot threads (greptile/cubic/codex) confirmed-and-fixed or answered, replied + resolved; CI green on that head. ⚠ observer-provider-injection.md is now only PARTIALLY fixed out-of-batch (JSONL names done; the broken `from pipelex import Pipelex` import at line 209 and pre-rename hook prose remain — follow-up stands, smaller). **Campaign branch reconciled with dev** (merge `4094fc050`, Phase-H style): 8 files conflicted (the two `CLAUDE.md`/`AGENTS.md` generated pair, kit agent_rules ×2, data-extraction, python-classes, PipeSequence, configure-ai-providers), all resolved by taking dev's reviewed superset — residual diff vs dev is campaign artifacts only (TODOS.md + `wip/drift-hunt/` + the kick-off stub). Earlier same day: **S2-4 DONE (2026-07-12)** (branch `docs/Drift-hunt-s2-4`, off dev `15e0a94ea`; campaign-branch mirror `5e15bec08` + records commit). All 23 features+tail findings fixed, **0 rejected-at-fix-time**; all cross-cutting sweeps landed (PipeStructure roster ×3 incl. the landing page — count-free phrasing per the no-hardcoded-counts rule; `agent-cli.md` `--mock-inputs` carry-over; D13 heartbeat everywhere). **+F11 (features file) and F14–F22 (tail file) fix-time drift found & fixed (per D16, adversarially verified), 3 candidates refuted** (Linkup/pypdfium2 extract-list "omissions" — correctly SCOPED lists, the scope qualifier is part of the claim; the debugging-page marker expression — provably selects the identical test set). **Two lessons with teeth:** (1) repo-root `CLAUDE.md`/`AGENTS.md` are GENERATED from `pipelex/kit/agent_rules/` — a direct edit failed `make check-rules` mid-batch; the D13 fixes were applied at the source templates + `make rules` (AGENTS.md picked up the fixes for free). (2) The batch's own new `LocalObserver` snippet was written with a broken import (`from pipelex import Pipelex`; real: `from pipelex.pipelex import Pipelex`) — caught pre-commit by import-checking every new snippet (the S2-1 lesson applied). The workspace-root `CLAUDE.md:160` heartbeat claim was fixed in ITS OWN repo (single-file commit `cc2b68c` on `feature/Follow-ups`). Empirical re-verification throughout (`pipelex init config` re-run sandboxed; the `uv` commands run live; pytest `-k` semantics proven by collection). 1 new code-side deferral (`deferred-code-findings.md` item 10: shipped `plxt.toml` "built-in default" comment contradicts its own `array_auto_collapse` value). 1 out-of-batch record for Louis: **`docs/advanced/observer-provider-injection.md` (+ `advanced/index.md:18`) still carry pre-rename observer hooks, wrong JSONL filenames, and the broken `Pipelex` import** — survived Part 2's stub classification and S2-3; one-page follow-up. Gates green (docs-check strict + full `make check`). **Blind post-fix review (2 no-context Sonnet reviewers on `66b1d6ebd`): ZERO defects.** `docs/setup/telemetry.md` untouched as ordered. Resolution records: banners + end sections of `findings/features-reliability.md` and `findings/the-tail.md`. Burn: ~1.5M tokens (13 fix + 6 verify + 2 review agents). ⟶ **NEXT = CHECKPOINT 2 (campaign close-out — STOP for Louis): full `make agent-test` on final state, `findings/SUMMARY.md` close-out (incl. the cookbook-unswept D12 disclosure + contract-candidate shortlist), hand off `deferred-code-findings.md` (10 items), prune `wip/drift-hunt/`.** Earlier: **S2-2 DONE (2026-07-12), PR #1044 → dev OPEN** (branch `docs/Drift-hunt-s2-2`, off dev, docs-only, 10 pages; campaign-branch commits `8f03eb6f2` + `1824f2db4`). All 14 concepts + top-level findings fixed, **0 rejected-at-fix-time**; **+7 fix-time drift found & fixed (F55–F61, per D16, adversarially verified) + 1 stale section pointer, and 6 candidates refuted** — notably the multi-line inline TOML tables candidate (the shipped tomli 2.4.1 accepts TOML 1.1; testing against the ACTUAL loader, not a generic parser, decided it) and the domain "three properties" claim (`main_pipe` is bundle/crate metadata the runtime `Domain` drops). Every changed example re-extracted from the pages and re-validated (cross-bundle via `-L` scaffolds); the libraries.md `execute()` pattern and the F3 Document Extraction sequence executed live in dry-run. **Blind post-fix review (2 no-context Sonnet reviewers on `8f03eb6f2`): ZERO defects — first Stage-2 batch to survive clean.** 2 new code-side deferrals (`deferred-code-findings.md` items 8–9: `tomli>=2.3.0` floor predates the TOML 1.1 syntax the docs/kit rely on; `build structures` opaque-placeholder shadowing after a class migration). 1 out-of-batch defect recorded for Louis: **`pipes/pipe-operators/PipeLLM.md:126-135` inline `@document` survived S2-1** (adversarially confirmed; one-line fix, could ride PR #1043's branch). Convention proposal for Louis (D11-adjacent): elided sequence sketches are elision-by-design iff the snippet validates once placeholder sub-pipes are scaffolded. Also: the globally-installed `pipelex-agent` shim backing the marketplace validate-mthds hook is broken in this workspace (`ModuleNotFoundError: pipelex_tools`) — hit independently by multiple agents. The PR carries this page set's unshipped Stage-0 fixes (native-concepts triple-quote + adapt-to-llm-prompting-style), self-contained on dev. Gates green (docs-check strict on both branches; `make check` green — check-urls flaked once, passed on retry). Resolution record: banner + end sections of `findings/building-methods.md`. Burn: ~1.5M tokens (8 fix + 4 verify + 2 review agents). ⟶ **NEXT = S2-4 (the last part), ⛔ needs Louis' per-part greenlight (D18).** Earlier: **S2-1 DONE (2026-07-12), PR #1043 → dev OPEN** (branch `docs/Drift-hunt-s2-1`, off dev, docs-only, 17 pages; campaign-branch commits `7942339cc` + `dd6174ee6`). All 30 `pipes/` findings fixed, **0 rejected-at-fix-time**; **+10 fix-time drift found & fixed (F45–F54, per D16, adversarially verified) and 2 candidates refuted** (the PipeCondition fallback fragment — fragment standard; the PipeExtract Overview one-liner — the code's own taxonomy covers web pages under "document processing"). Every changed bundle example re-validated against the real parser incl. dry-run; the blind post-fix review caught **2 defects in the batch's own text** — the "run a bundle" Python snippets on TWO pages never ran (`bundle_uris` is only a dedup key; `execute()` needs `pipe_code` or `mthds_contents` — inherited from the original docs, the Stage-1 kwarg-rename finding masked the broken call pattern) and the F45 dedup claim missed its `bundle_uris` gate. Corrected snippets executed live in dry-run. 2 new code-side deferrals (`deferred-code-findings.md` items 6–7: CSV codec datetime inconsistency; concept-factory `refines` asymmetry). Gates green (docs-check strict on both branches + `make check`). Resolution record: banner + end sections of `findings/building-methods.md`. Burn: ~1.7M tokens (10 fix + 3 verify + 2 review agents). ⚠ S2-2 lesson: **a Stage-1 finding scoped to a kwarg can hide a non-functional surrounding example — validate the whole snippet, not just the finding's clause.** ⟶ **NEXT = S2-2 (recommended), ⛔ needs Louis' per-part greenlight (D18).** Earlier: **S2-3 DONE (2026-07-12), PR #1042 → dev OPEN** (branch `docs/Drift-hunt-s2-3`, off dev, docs-only). All 42 under-the-hood+advanced findings fixed, **0 rejected-at-fix-time**; +4 fix-time drift found & fixed (F43–F46, per D16, adversarially verified) and **2 fix-time candidates REFUTED — the campaign's first refutations** (fix-time finders over-flag; the D6 verify gate is what keeps D16 honest). 2 code-side items deferred (`deferred-code-findings.md` items 4–5), F24 fixed at its real site + class label corrected, `setup/telemetry.md` untouched. The blind post-fix review (2 no-context Sonnet reviewers on the commit SHA) caught 3 defects in the batch's own new text — including one **inherited from the Stage-1 finding's Reality text itself** (F36's "packaged kit configs" fallback; real fallback is global `~/.pipelex/`) — all folded in before the PR. ⚠ Lesson for S2-1/2/4: D6's "code is the authority" applies to the finding's *Reality* text too, and the blind post-fix review earns its keep — keep both. Gates green (docs-check strict on both branches + `make check`). Resolution record: banner + end section of `findings/under-the-hood-advanced.md`. Burn: ~2.4M tokens (14 fix + 2 verify + 2 review agents). ⟶ **NEXT = S2-1 (recommended), ⛔ needs Louis' per-part greenlight (D18).** Earlier: **🎉 STAGE 1 COMPLETE — PART 4 DONE (2026-07-12).** The tail (18 pages: get-started 3, contribute 3, tools non-CLI 2, setup 1, viewpoint 1, root index/license 2, D9 additions 6) swept via the wave Workflow (`wf_0c7af057-a6e`): 244 claims checked, **12 confirmed / 0 rejected**, **+1 more found by the post-run false-negative check** = **13 drift**, 0 reclassified. Severity: 2 breaks-a-user / 9 wrong / 2 misleading / 0 stale; **density 0.72 (9/18 dirty)**. Findings in `wip/drift-hunt/findings/the-tail.md`; `SUMMARY.md` now carries the closed Stage-1 scoreboard. No fixes applied (fixes = Stage 2). **STAGE 1 TOTAL: 104 pages, 1,439 claims, 109 drift, 3 convention, 0 rejected, density 1.05.** ⟶ **NEXT = CHECKPOINT 1 (STOP for Louis).** Earlier: Parts 1–3 recaps below; CHECKPOINT 0 CLOSED (D7–D10), Phase H DONE, Phase B DONE, Stage 0 DONE, gates green.
+- **Part 4 recap + what it taught (read before Stage 2):** the tail is mostly **contributor / process / meta** pages, so its dominant claim type was one the campaign had never met — **repo tooling** (make targets, generated file trees, `mkdocs.yml`, `robots.txt`, the Python floor). Four new review rules were added on top of the reused Part-3 block: **H** make-targets-and-tooling (grep the Makefile *and read the recipe*), **I** process/policy prose is out of scope (zero findings on a code-of-conduct is the *correct* outcome), **J** version requirements (`requires-python = ">=3.11,<3.15"`; 3.10 dropped — fired zero times, `CONTRIBUTING.md` is correct), **K** docs-site meta-claims (vs `mkdocs.yml` / workflows / Makefile). **Three headline results:** (1) **the drift correlates with audience, not complexity** — every maintainer-facing page came back clean (drift-contracts, keyword-only, telemetry setup), every page explaining tooling *to someone else* drifted; (2) **docs edited in the same commit as their code do not drift** (proven: `d3550330a`, `15e0a94ea` each touched doc+code together) — direct evidence for the drift-contracts mechanism; (3) **the tooling class wants a CONTRACT, not a check** — every documented make target *exists*, so an existence script scores zero; what drifted is *behavior under a stable name* (`agent-test` grew a heartbeat 3.5 weeks after the page was written on the premise it was silent).
+- **⚠ False-negative discipline — IT FINALLY CAUGHT SOMETHING (Part 4).** The rule (a *clean* result on a page seeded as high-risk is where the campaign most likely fools itself) held in Parts 1–3 every time. In Part 4 it **broke a clean verdict**: the language tutorial's four bundles all pass `pipelex validate bundle` *and* dry-run — and the page still teaches a **false rule** about the `@` vs `$` sigils (F13: it claims they differ by *data provenance*, input-vs-working-memory; they actually differ by *rendering shape*, block-on-its-own-line vs inline, and both resolve against the same `declared_inputs`). **Lesson with teeth for Stage 2 and for the contract design: a snippet validator verifies the snippet, never the sentence beside it.** With 109 confirmed and **0 ever rejected**, the pipeline's error mode is proven to be **false negatives, not false positives** — so keep the FN spot-check as a standing step.
+- **⚠ Correction to the campaign's own record (fix at Stage 2):** Part 2's **F24 is right as written** (`pipelex-agent init --config` doesn't accept a `telemetry_mode` field) but its **class label — "fabricated `telemetry_mode` values" — is WRONG**. The values are real (`PostHogMode` StrEnum = `off`/`anonymous`/`identified`, `telemetry_config.py:27-32`) and `docs/setup/telemetry.md` documents them **correctly** (it FN-checked clean, decisively). A Stage-2 fixer trusting the class label could gut an accurate page. Fix F24 at `init-cli-flows.md:404`; leave `setup/telemetry.md` alone; correct the label in `under-the-hood-advanced.md`. **✅ DONE in S2-3 (all three parts honored).**
+- **Part 2 recap (DONE 2026-07-12):** under-the-hood (19) + advanced (7) = 26 pages, 435 claims, 42 confirmed / 0 rejected / 0 reclassified (10 breaks / 23 wrong / 4 misleading / 5 stale; density 1.62, or 2.0 vs the 21 substantive pages — 5 `advanced/*-injection.md` are Under-Construction stubs). Findings in `wip/drift-hunt/findings/under-the-hood-advanced.md`. Dominant class: renamed Python API never propagated into docs.
+- **Part 1 recap (DONE 2026-07-12):** building-methods (35 pages), 452 claims, 47 confirmed / 0 rejected → **3 reclassified as authoring convention (D11) → 44 drift** (25 breaks / 11 wrong / 6 misleading / 2 stale; density 1.26). Findings in `wip/drift-hunt/findings/building-methods.md`.
+- **Next:** **NOTHING — campaign closed at Checkpoint 2 (2026-07-12).** Awaiting Louis' final review of the close-out. Campaign branch: do NOT push unless Louis asks (origin copy is stale at `2cb9c5957`, branch is ahead — expected). Post-campaign follow-ups live outside this tracker: the drift-contracts Phase 3 verdict consumes the shortlist in `findings/SUMMARY.md`; the 10 deferred code findings are ordinary follow-ups from `deferred-code-findings.md`.
+- **Carry-overs into Stage 2:** ✅ ALL DONE in S2-4 — `docs/tools/cli/agent-cli.md:38` fixed; the `PipeStructure` roster fixed on all 3 pages (incl. a new PipeStructure section on `features/pipe-operators.md`). (The F24 class-label correction was ✅ done in S2-3.)
+- **Reusable wave script (for Stage C, the last fan-out):** `/Users/lchoquel/.claude/projects/-Users-lchoquel-repos-Pipelex--drift/268d8cee-6e55-4669-9dff-33187cf92158/workflows/scripts/drift-hunt-tail-part4-wf_0c7af057-a6e.js` — start from **this** one (not Part 3's): it carries the full rule stack (Part-3's A–G + Part-4's H–L) and the D11 carve-out. Swap `PAGES` + seeds. For Stage C, add the D3 ruling (cookbook ground truth is `../pipelex-cookbook`, resolve references there) and **budget for the FN spot-check as a first-class step, not an afterthought**.
+- **Phase H carved out (2026-07-12, Louis-approved):** the drift-tool hardening shipped separately as **PR #1041 → dev MERGED** (`15e0a94ea`, branch `fix/Drift-ack-hardening`). **dev merged into this branch 2026-07-12** (`19f991300`): the 4 drift-ack files conflicted (not the "content-identical, no conflict" originally predicted — #1041 landed as a *reviewed superset* of the branch's early `8c7413afc`, adding the post-verify recheck + staging-failure rollback), all resolved by taking dev's superset, so the drift area is now byte-identical to dev. Gates green (agent-check clean, `test_drift_ack_cmd.py` 18 passed incl. #1041's 2 new tests). Branch is now 0 behind dev.
+- **Branch / worktree:** `docs/Drift-hunt` off `dev`, in the `_drift` worktree (treat as repo root). Do NOT push unless Louis asks. Commits so far: tracker `5e3699c64`, hardening `8c7413afc`, backfill `4093b1628`, then Stage 0 artifacts + fixes, then **dev merge `19f991300`** + a `ruff-format` normalization `86b569acf`, then **Stage 1 Part 1 findings** (see git log).
+- **Working artifacts (post-prune):** `wip/drift-hunt/` — `findings/SUMMARY.md` (the campaign record + close-out), `inventory.md` (the denominator), `deferred-code-findings.md` (the handoff). Everything else pruned at Checkpoint 2; recover via `git show drift-hunt-archive:wip/drift-hunt/<path>` (pushed tag pinning `9a6201a5e`).
+- **Decisions taken:** D1–D10 below, plus H-A2 implemented (narrow warn→fail escalation — see Phase H).
+- **Key learning (D6 vindicated):** `pipelex validate --all` was initially "confirmed dead" from `--help` (no group-level `--all`) and a fix was drafted — live execution proved a shortcut layer forwards it to `validate pipe`; fix reverted. Never conclude a CLI form is dead from `--help` alone; run it.
+- **Open questions:** none — all rulings taken (D1–D19), campaign closed. The only pending act is Louis' final review.
-### Scan snapshot (2026-07-10 — point-in-time measurements, they WILL drift; regenerate via the Phase 1 `--report` extension)
+## Decisions
-- ~2,600 defs inspected by the guard (after carve-outs); ~650 of them take no params.
-- **~1,700 use Exception 1 today** (~1,000 lone-subject, ~700 subject-plus-kwonly) — roughly 9 in 10 param-bearing inspected defs. The stock is the house style and looks overwhelmingly legitimate (top subject names: `name`, `pipe_code`, `exc`, `value`, `path`, `content`, `data`, `node`, …) — which is why the stock is seeded-then-reviewed rather than mass-demoted.
-- ~220 defs are already fully keyword-only; zero use `/`; a handful of `# kw-only: ignore` hatches.
-- **~30 literal-typed subjects** (`bool`/`int`/`float`, e.g. `do_doctor_cmd(fix: bool)`) — the mechanically-catchable slice of exactly the abuse this project kills. Beware: some are framework-positional (`version_callback(value: bool)` is invoked positionally by Typer via `callback=` — pyright is blind to that; `make agent-test` is the safety net).
-
-### Key repo facts (verified 2026-07-10)
-
-- Guard core: `pipelex/cli/dev_cli/commands/keyword_only_guard.py` (pure stdlib; `_evaluate_def` order = dunder → escape hatch → decorator carve-outs → typer-annotation → symmetric allowlist → rule). Presentation layer: `check_keyword_only_cmd.py`. Hook: `.claude/hooks/check-keyword-only.sh` invokes the guard **by file path** (never `-m` — that would import the `pipelex` package chain and blow the cold-start budget).
-- `make fko` inserts the bare `*` leftmost (all-keyword form) and is non-gating; read-only `make cko` owns the gate and runs last in `agent-check`, in `make check`, and as the `lint-keyword-only` CI job (aggregated by `lint-all`).
-- Convention doc: `docs/contribute/keyword-only-arguments.md`. **Drift contract `keyword-only-convention`** (in root `drift.toml`) triggers on the guard file with that doc as review target → Phase 3 must stage the trigger and `make drift-ack CONTRACT=keyword-only-convention RATIONALE="…"` after genuinely updating the doc. The new dev-CLI command sits under `pipelex/cli/dev_cli/**`, which the `cli-docs` contract excludes — only the one contract opens.
-- Dev CLI house style: Typer wrapper in `_dev_cli.py` delegating to a keyword-only `*_cmd()` in the module; CI-gate idiom prints a rich panel then `sys.exit(1)` (see `check_config_sync_cmd.py`). Unit tests: `tests/unit/pipelex/cli/dev/`, inline-source style (`find_violations_in_source` on snippets); a reusable `git_repo` fixture already exists in that conftest (built by the drift project) if needed.
-- If new CLI error classes are added (grant-command failures), they go in an `exceptions.py` per house rules and require `make gep` (error-pages regeneration) in the same commit.
-- Agent-rules source: the keyword-only section of `.claude/rules/python-standards.md` (workspace root) is generated from the kit rules source in this repo — edit the kit source, then regenerate (`make rules` family), never the generated file.
-- Makefile: per-target + shorthand-alias pattern (`check-keyword-only`/`cko`); `.PHONY` is one block. mkdocs nav lists contribute docs in TWO places in `mkdocs.yml` — update both if a doc is added (none planned; the convention doc already exists).
-
-## Checkpoint protocol (mandatory at every CHECKPOINT below)
-
-1. **Verify:** the phase's gates green — `make agent-check`, full `make agent-test` (targeted tests are fine between checkpoints), plus the checkpoint's specific gates. Do not proceed with failures.
-2. **Commit** the phase's work as one coherent commit (do not push unless asked).
-3. **Update this file** — tick boxes, refresh Cold-start context (status, decisions, deviations, commit SHAs) so a brand-new session resumes with zero conversation context.
-4. **Fan out `/code-review`** — a fresh no-context Sonnet sub-agent (Agent tool, `general-purpose`, `model: sonnet`, never a fork) pointed only at the commit range; never hand it this plan or your own conclusions.
-5. **Triage findings:** fix real defects; findings that are design tradeoffs get captured in a deferred-items doc under `wip/subject-grants/`, not reflexively applied.
-6. **STOP** — natural handoff point.
-
-## Rubric — what earns a grant
-
-Applies to every Phase 4 review and every future grant. Formalized into the convention doc in Phase 3; until then this section is the reference.
-
-A subject grant is warranted when EITHER:
-
-- **Verb–object test:** the function name is a verb (phrase) and the param is its direct object — the call reads as a sentence (`render(node)`, `validate_bundle(bundle)`, `parse_concept_spec(spec_data)`) — AND it is the **single candidate** (if you hesitate between two params, neither is the subject) — AND typical call sites pass a **self-labelling expression**, never a bare literal (literal-typed subjects are banned by D4 anyway);
-- OR the def must satisfy a **positional `Callable` protocol** (it is passed as a value to something that calls it positionally) and a grant keeps it compliant without reaching for the heavier `# kw-only: ignore`.
-
-When in doubt → keyword-only. The grant is the exception tier; all-keyword is always compliant and often more readable. Rationales must be def-specific but may be terse for obvious keeps ("verb–object; single operand") — the value is that someone actually looked; copy-paste boilerplate across dozens of entries defeats the point and is what Louis will spot-check for.
-
-## Registry format
-
-```toml
-version = 1
-
-["pipelex/graph/render.py::render_node"]
-param = "node"
-rationale = "Verb–object: renders the node; single obvious operand."
-
-# Transitional shape during Phases 2-4 only (rejected entirely by the guard after Phase 5):
-["pipelex/foo/bar.py::SomeClass.some_method"]
-param = "spec"
-rationale = "SEEDED 2026-07 — pre-registry tree, treated as new, review pending"
-seeded = true
-```
-
-Schema after finalization: exactly `param` + `rationale` per entry, `version` at top; unknown keys are a check failure.
-
-## Phase 1 — Mechanism: guard + registry + commands
-
-- [x] Guard core (`keyword_only_guard.py`, stays stdlib-only): load `subject_grants.toml` from repo root (`tomllib`); missing registry file = explicit check error (not mass-violation). New rule: a positional-or-keyword subject requires a matching grant (key + `param`); a literal-typed subject (D4) is a violation even with a grant; grants with no matching def or mismatched `param` are check failures. `# kw-only: ignore` keeps short-circuiting everything (existing evaluation order).
-- [x] Distinct violation messages per kind — "ungranted positional subject" / "banned literal-typed subject" / "stale grant" — each naming its fix (`make fko`, `make subject-grant FUNC=… RATIONALE=…`, or registry cleanup). Update the lean `main()` hook message text too.
-- [x] `pipelex-dev subject-grant "<path>::<qualname>" --rationale "…"`: validates the def exists, its first non-`self`/`cls` param is positional-or-keyword and not literal-typed, rationale non-empty; records `param` automatically; rewrites the registry sorted. Make target `subject-grant` (+ short alias, e.g. `sgr`), following the existing per-target pattern.
-- [x] Temporary `--seed` mode on `subject-grant` (deleted in Phase 5): scan the tree, emit an entry for every existing positional subject with `seeded = true` + the placeholder rationale.
-- [x] Extend `check-keyword-only --report`: per-package grant totals and seeded-remaining counts — this is the Phase 4 progress meter.
-- [x] Verify `fko` handles the new violation kind (it already inserts `*` leftmost; ungranted subject defs must be mechanically fixable; the re-parse-before-write guarantee is unchanged).
-- [x] Unit tests (inline-source style + injected registry content): grant match / missing grant / param mismatch / dead entry / seeded accepted (transitional) / literal ban incl. union and Optional forms / ignore-hatch precedence / same-qualname sharing (D11) / lean single-file path reads the registry / sorted-write round-trip / grant-command refusals (literal subject, non-positional subject, empty rationale, missing def).
-- [x] New code itself passes the convention; new error classes (if any) in `exceptions.py` + `make gep`.
-
-## Phase 2 — Seed + literal-subject sweep (lands atomically with Phase 1 — the tree must never gate red)
-
-⚠ **Ordering hazard:** between flipping the guard and committing the seed, NEVER run `make fko` / `make agent-check` on the tree — the fixer would kwonly ~1,700 defs. Generate the seed first (standalone scan), commit guard + registry together.
+- **D1 — cocode is out of scope.** Not ready for this task (user call, 2026-07-05). The whole hunt runs on the pre-screen script + Claude agent fan-outs; do not reach for cocode even for sub-steps.
+- **D2 — scope = hand-written pages under `docs/`.** Excluded as *generated*: `docs/errors/` (regenerated by `make gep`) and the generated gateway-models doc(s). Excluded as *freshly reviewed*: `docs/configuration/` and the CLI docs (`docs/tools/cli/` + `pipelex/cli/agent_cli/CLAUDE.md`) — they got a genuine review as the drift-contracts seed acks; re-sweeping them now would measure nothing. Everything else under `docs/` is in: building-methods, features, cookbook, under-the-hood, tools (non-CLI), advanced, reliability, get-started, contribute, setup, viewpoint, index/license.
+- **D3 — cookbook ground truth is a sibling repo.** `docs/cookbook/` pages reference examples that live in `../pipelex-cookbook` (read-only ground truth), not in this tree. Reviewers and the pre-screen's path-existence check must resolve cookbook references there; never conclude "referenced example missing" from this tree alone.
+- **D4 — fixes land in reviewable batches.** One commit per docs section at minimum; separate PRs per section if a batch grows beyond comfortable review size. Default base: `dev`.
+- **D5 — no manifest growth during the hunt.** The drift-contracts dogfood freeze stays in force: the hunt produces *evidence* for which sections might earn a contract later; it does not add contracts to `drift.toml`.
+- **D6 — findings must survive adversarial verification before anything is fixed.** No finding is acted on unless a second, independent agent confirms it against the code with `file:line` evidence. Fix agents re-verify at fix time; a finding that doesn't hold up gets recorded as rejected, not silently dropped.
+- **D7 — cookbook cross-check is its own greenlight-gated stage (Louis, 2026-07-12).** Comparing `docs/cookbook/` pages against `../pipelex-cookbook`'s actual files runs as Stage C, separate from Stage 1 and ONLY on Louis' explicit greenlight — the cookbook may change soon. Cookbook pages are out of the Stage 1 fleet.
+- **D8 — cross-repo references: minimal only; code defects get deferred, not fixed (Louis, 2026-07-12).** A short, named pointer to another repo's own docs is fine; workspace-relative paths a docs-site reader can't resolve, or restating another repo's content at length, is drift — inline the short essence here, leave the detail in its home repo. Code defects surfaced by the hunt are not fixed mid-campaign: record them in `wip/drift-hunt/deferred-code-findings.md`.
+- **D9 — scope edges: all in (Louis, 2026-07-12).** The four pages D2 didn't name (`docs/agents/debugging-hanging-pytest-runs.md`, `docs/analytics/data-extraction.md`, `docs/distributed-execution/index.md`, `docs/CLAUDE.md`) AND the root-level `CONTRIBUTING.md` + `CODE_OF_CONDUCT.md` (the include-targets of the docs stubs) join Stage 1. Note: the two root files got no Stage 0 pre-screen (it scanned `docs/` only) — their review agents start without seed rows.
+- **D11 — scaffolding omissions in teaching examples are authoring convention, not drift (Louis, 2026-07-12).** Omitting a top-level `domain`, a pipe `description`, or the `prompt` body in a focused example is NOT drift — a reader fills it in (with a placeholder) before running. **Still drift** (survives scaffolding): a written prompt that ignores its own declared input; a functionally-required input that IS the subject (e.g. PipeExtract's Image/Document); wrong field names / output types / enum values / defaults. Applies to ALL Stage-1 parts and Stage C — **seed every future review agent with this carve-out** alongside the FP do-not-flag list. Consequence for the derived-check candidate: `validate-doc-bundle-examples` must inject placeholder `description`/`domain` and tolerate an omitted prompt before running `pipelex validate bundle`, so it only flags defects that survive scaffolding.
+- **D12 — Stage C (cookbook) is DEFERRED (Louis, 2026-07-12).** Not part of this campaign's remaining arc; supersedes D7's greenlight gate. The 21 cookbook pages stay unswept — say so plainly in the close-out (Stage 1 covered 104 of the 125 in-scope pages).
+- **D13 — the `make agent-test` heartbeat claim gets fixed EVERYWHERE, in Stage 2 (Louis, 2026-07-12).** `agent-test` is no longer silent on success (start line + 20 s heartbeat + completion line — `Makefile:732-739`, `:49-65`, `:27`). Fix Part-4 F8's doc page **and** the two agent-facing instruction files that are its source: `CLAUDE.md:52` ("Zero output on success") and the **workspace-root `CLAUDE.md:160`** ("silent on success"). ⚠ **The workspace-root file is in a DIFFERENT GIT REPO** (`/Users/lchoquel/repos/Pipelex` is its own repo; this worktree is `pipelex`) — it cannot ride this branch. It needs its own commit over there, in the same S2-4 session.
+- **D14 — Stage 2 = 4 parts, ONE PR PER PART to `dev`, one cold-start session each (Louis, 2026-07-12).** Independent, not stacked: the parts touch disjoint page sets, so they can land in any order. Each part resumes cold from its own findings file alone.
+- **D15 — fixers MAY restructure a broken teaching example (Louis, 2026-07-12).** When a correct fix needs more than a one-line patch (e.g. a prompt that ignores its own declared input), the fixer may rewrite the example so it is correct and actually runs. **Bounded:** no new content, no editorial rewrites, and the example must **re-validate against the real parser** before the fix counts. This is what many of the ~18 building-methods bundle-example findings actually require.
+- **D16 — drift spotted DURING a fix: fix in place, if it clears the same bar (Louis, 2026-07-12).** Fix agents re-verify against the code (D6), so they will find drift Stage 1 missed — F13 proves the sweep is not exhaustive. If the new drift is **on a page already in the batch** and clears the same evidence bar (`file:line` on both sides, adversarially verified), fix it in the same batch and **append it to the findings file** (the audit trail must stay complete). Drift on an out-of-scope page is **recorded, not fixed**.
+- **D17 — Stage 2 stays DOCS-ONLY (Louis, 2026-07-12).** No derived checks, no new `drift.toml` contracts (D5 holds). The contract/check shortlist is handed to the drift-contracts Phase 3 verdict at Checkpoint 2. Keeps every Stage-2 PR pure-docs and the campaign's evidence honest — and Part 4 is the cautionary tale: a green snippet check reads as a clean page when it is not (F13).
+- **D18 — ⛔ NO Stage-2 part starts without Louis' explicit greenlight (Louis, 2026-07-12).** Same gate as D10 governed Stage 1: per part, he says go. Do not begin fixing because the plan looks settled.
+- **D19 — elided sequence sketches are authoring convention, not drift (Louis, 2026-07-12).** A sequence sketch whose step pipes are absent from the snippet is elision-by-design **iff the snippet validates once placeholder sub-pipes are scaffolded** (and no prose claims it is complete or runnable). Ruled during S2-2 (the native-concepts "In Complex Methods" refutation; same pattern at `working-memory.md:24-27`). D11-adjacent: seed every future review/fix/verify agent with this carve-out alongside D11's, so the pattern is ruled uniformly instead of re-litigated per page. Still drift: a sketch that fails even with placeholder sub-pipes scaffolded (wrong step syntax, bad concept refs, sequence-output mismatch).
+- **D10 — Stage 1 runs as 4 parts, one cold-start session each (Louis, 2026-07-12).** Full fan-out, worst-first, but split so context is cleared between parts and each part resumes cold from this tracker: Part 1 building-methods (35) → Part 2 under-the-hood + advanced (26) → Part 3 features + reliability (25) → Part 4 the tail: get-started, contribute, tools non-CLI, setup, viewpoint, root index/license, the D9 additions (18). Per part: run the wave Workflow, commit `findings/`, update Cold-start context (findings count + token burn), report to Louis — he greenlights the next part (or stops) before a new session picks it up.
-- [x] Run `--seed`, commit the registry: every existing positional subject enters as `seeded = true`.
-- [x] Sweep the literal-typed subjects (~30 per snapshot): kwonly each (fix call sites, pyright-guided), EXCEPT framework-positional ones (Typer `callback=`, etc.) which get `# kw-only: ignore` with a nearby justification. Remove their seed entries. `make agent-test` is the safety net for the pyright-blind cases.
-- [x] Gates green: `make agent-check`, `make agent-test`, `make cko`.
+## What "drift" means here (and what it doesn't)
-## Phase 3 — Docs, rules, drift-ack (before the grind — mid-grind sessions must read the NEW convention, not the stale one)
+In scope per page: statements the code can contradict — commands, flags, config keys, defaults, file paths, API/class/function names, described behaviors and sequences, code snippets that should run or parse. Out of scope: editorial quality, tone, structure, SEO, completeness wishes. This is a drift hunt, not a rewrite; standing repo rules apply to fixes (docs = current reality, no hardcoded counts, MkDocs conventions).
-- [x] Rewrite `docs/contribute/keyword-only-arguments.md`: Exception 1 becomes "granted subjects" (registry, commands, staleness semantics, D4 literal ban), add the rubric verbatim, record the reversal honestly (generic permission → explicit grant, and why: agents follow checks, not prose). Keep the no-`/` decision section (D9).
-- [x] Update the kit rules source for the keyword-only section (the generated `.claude/rules/python-standards.md` block) + regenerate; update the summary block in this repo's `CLAUDE.md`.
-- [x] Drift: stage the trigger files, then `make drift-ack CONTRACT=keyword-only-convention RATIONALE="…"` — a genuine review, and a real dogfood data point for the drift system (log it via the drift-review skill's dogfood log).
+## Checkpoint protocol (applies to every checkpoint below)
-### CHECKPOINT 1 — mechanism live, seeded, documented
+1. **Gates:** `make agent-check` (any code change), `make docs-check` (mkdocs strict — any doc change), full `make agent-test` at checkpoints.
+2. **Commit** the phase's work (per D4 granularity for doc fixes). No push.
+3. **Update the Cold-start context** section above (status, commit SHAs, decisions, open questions) so a fresh session can resume cold.
+4. **Fan out `/code-review`** on the phase's changes: Sonnet-5 sub-agent(s), spawned with **no inherited context** — hand each one only a pointer to the changes (commit SHA or `git diff <base>..HEAD` range), never the plan, the rationale, or our own conclusions. Triage findings for clean-solid-not-over-engineered; apply what survives.
+5. **STOP for Louis** where marked — Checkpoints 0, 1, and 2 are user-review gates.
-Everything except the grind is done and reviewed as one unit. Gates: full checkpoint protocol + `make check` end-to-end (includes `drift-check`) + mkdocs strict build if docs changed.
+## Deferred drift-contracts items (track here, do NOT do during the hunt)
-## Phase 4 — The review grind (every seeded grant treated as new)
+- [ ] **Phase 3 — pilot verdict:** per-contract keep / narrow / mechanize / drop decision, driven by `wip/drift-contracts/dogfood-log.md`. Gated on the dogfood period verdict; the hunt feeds it evidence (Stage 1 logs observations; Checkpoint 2 hands over the contract-candidate shortlist). **Working doc drafted 2026-07-13: `wip/drift-contracts/phase-3-verdict.md` — awaiting Louis' rulings on its decision boxes.** (Original phase breakdown: the `docs/Update` branch is gone; the surviving source is `wip/drift-contracts-design.md` → Rollout plan.)
+- [ ] **lint-drift promotion into lint-all / required CI check:** gated on the open-PR backlog landing (a required check would fail every open PR that predates its acks). Track only.
+- [x] **PR #1040 hardening findings (A1/A2, `wip/drift-contracts/pr-1040-review-notes.md`):** resolved in Phase H below.
-Batch by top-level package (`pipelex/core`, `pipelex/builder`, `pipelex/cli`, `pipelex/cogt`, …), sized via `--report` — expect on the order of 10–15 batches for ~1,700 entries. This phase is the dominant cost of the project.
+---
-Per seeded entry, apply the rubric and pick exactly one:
+## Phase H — hardening pass (drift tool, from PR #1040 review notes)
-- **Keep:** replace the placeholder with a real, def-specific rationale; drop `seeded`.
-- **Demote:** convert the def to all-keyword (`*` leftmost), fix call sites (pyright-guided), delete the entry.
-- **Hatch (rare):** `# kw-only: ignore` when the positional shape is framework-owned; delete the entry.
+- [x] **H-A1 — `drift ack` auto-stages the ack file it writes.** `git add .drift/acks/<id>.toml` right after `save_ack`, via a new `stage_file` git-adapter function (subject grant recorded BEFORE running checks). Makes the local gate honest: the ack lands in the same index `drift check` reads. Temp-repo test: ack path appears in `git diff --cached --name-only` after `drift ack`.
+- [x] **H-A2 — DECISION: implement the narrow warn→fail escalation.** When a contract has `verify_commands` AND an unstaged-modified or untracked file matches its triggers, `drift ack` fails instead of warning — that is the one case where the verify guarantee is meaningfully weakened (verify ran on working-tree content the digest does not cover). Check runs BEFORE the verify commands (fail fast, don't burn a `make tb`). No-verify contracts keep the warn-only behavior. Tests: verify-contract + dirty trigger → `DriftAckError`, ack not written, verify not run; no-verify contract keeps warning + writing.
+- [x] Update `docs/contribute/drift-contracts.md` + `.claude/skills/drift-review/SKILL.md` (ack now auto-stages; warn→fail rule) + `CHANGELOG.md` [Unreleased].
+- [x] Gates: `make agent-check`, targeted tests (`tests/unit/pipelex/cli/dev/`), `make docs-check`.
-Rules of engagement:
+## Phase B — dogfood-log backfill
-- [x] Each batch is one commit, gates green (`make agent-check` + targeted tests; full `make agent-test` at checkpoints and after any batch that demoted framework-adjacent defs). — done for all 11 batches; full agent-test run after each of batches 8–11.
-- [x] Expect drift toward all-keyword — that is desired; grants are the luxury tier. — confirmed: ~150 demoted, ~1,453 kept overall.
-- [x] **Demoted public surfaces:** appended below (see the new batch-8–11 entries). No backward compat, but the breakage is inventoried for the release-wave cross-repo sweep.
+- [x] Backfill the missing mandatory entry in `wip/drift-contracts/dogfood-log.md` for the `config-docs` ack recorded 2026-07-11 during the subject-grants regrind (ack rationale: Python-3.10 drop / StrEnum swap in `img_gen_job_components.py` via the Phase-M dev merge — clean-pass). Mark it honestly as a backfill.
-Demoted public surfaces (append as you grind — the Phase 2 literal sweep already demoted these; external consumers calling them positionally break):
+### CHECKPOINT H — verify + review (no user stop)
-- `pipelex.tools.misc.string_utils.pluralize` / `count_with_noun` — `count` now keyword-only
-- `pipelex.tools.log.log.Log.set_level_by_int` (`level_int`) and `pipelex.tools.log.log_levels.LogLevel.from_int` (`logging_level`)
-- `pipelex.tools.misc.pretty.PrettyPrinter.pretty_width` (`width`)
-- `pipelex.hub.PipelexHub.set_dry_run_forced` (`is_forced`)
-- `pipelex.cli.installed_methods.discover_installed_methods` (`include_global`)
-- `pipelex.cli.commands.init.config_files.init_config` (`reset`)
-- `pipelex.system.telemetry.*.is_custom_portkey_logging_enabled` / `is_pipelex_gateway_portkey_logging_enabled` (`is_debug_configured`; internal callers already keyword)
-- `pipelex.plugins.anthropic.anthropic_factory.AnthropicFactory.calculate_safe_max_tokens_for_timeout` (`timeout_seconds`)
-- `pipelex.runtime_bridge.bootstrap.ensure_pipelex_booted` (`config_overrides`) — runtime_bridge is consumed by pipelex-api / pipelex-worker; positional callers break
-- `pipelex.builder.operations` entry points `build_inputs_for_pipe` (`pipe_code`), `list_models` (`categories`), `validate_all` (`library_dirs`) — agent-CLI/MCP plumbing, but importable
-- `pipelex.errors.error_pages_generator.generate_error_pages` (`output_dir`) — dev tooling, listed for completeness
-- `pipelex.reporting` protocol methods `set_event_log` / `clear_event_log` (`context_key`) — external ReportingProtocol implementations must match
-- `pipelex.core.pipes.pipe_abstract.PipeAbstract` run-family (`run_pipe`, `live_run_pipe`, `dry_run_pipe`, `validate_before_run`, `validate_after_run`, `needed_inputs`) — `job_metadata` / `visited_pipes` now keyword-only; any external code subclassing or invoking pipes positionally breaks
-- `pipelex.core.stuffs.stuff_content.StuffContent.rendered_for_prompt` / `rendered_for_template_async` (`text_format`) and `pretty_print_content` (`title`) — public content API; positional TextFormat callers break
-- `pipelex.libraries.library_manager_abstract.LibraryManagerAbstract` load family (`load_libraries`, `load_from_blueprints`, `load_from_crate`, …) — `library_id` now keyword-only
-- `pipelex.pipeline.pipeline_run_setup.pipeline_run_setup` (`execution_config`) and `pipelex.pipeline.pipeline_manager_abstract.add_new_pipeline` (`pipe_code`)
-- `pipelex.tools.jinja2.image_renderable.ImageRenderable.render_with_images` (`registry`) — protocol + all content-class implementations
-- CLI-internal helpers (doctor/show/update/validate `*_cmd` delegates and UI builders) — almost certainly not called externally, listed for completeness in the diff, not here
-- `pipelex.tools.jinja2.text_format_renderable.TextFormatRenderable.rendered_for_template_async` (`text_format`) — `@runtime_checkable` protocol; external implementers/positional callers break (checkpoint-2 spillover completion)
-- `pipelex.libraries.concept.concept_library_abstract.ConceptLibraryAbstract.is_compatible` / `concept_library.ConceptLibrary.is_compatible` (`tested_concept`) — external code subclassing or calling `is_compatible` positionally breaks (checkpoint-2 demote)
-- **Batch 8–11 demoted public surfaces (external consumers calling positionally break):**
- - `pipelex.cogt.image.prompt_image_factory.PromptImageFactory.make_prompt_image` (`uri`) and `pipelex.cogt.document.prompt_document_factory.PromptDocumentFactory.make_prompt_document` (`uri`) — public content factories; positional-`uri` callers (cookbook, cocode, app) break
- - `pipelex.cogt.models.model_manager_abstract.ModelManagerAbstract.setup` + its `ModelManager.setup` override (`secrets_provider`) and `pipelex.cogt.model_backends.backend_library.InferenceBackendLibrary.load` (`secrets_provider`) — external code subclassing/invoking the model-manager setup/backend-load positionally breaks (dependency-first demote, matches `load_telemetry_config`)
- - `pipelex.cogt.llm.llm_setting.LLMSettingChoices.make_completed_with_defaults` (`for_text`) and `pipelex.cogt.model_routing.routing_profile.RoutingProfile.get_backend_match_for_model` (`enabled_backends`)
- - `pipelex.tools.aws.aws_config.AwsConfig.get_aws_access_keys_with_method` (`api_key_method`) — mode selector
- - The remaining batch-8–11 demotes are CLI-internal / plugin-internal helpers (list_*_models, discovery register helpers, console UI, doctor check_*, run/validate/build cores, gateway `_call_relay`) — not part of the importable `pipelex` public API; listed in the commit diffs, not here.
+- [x] Checkpoint protocol steps 1–4 (commit Phase H + B; `/code-review` fan-out on the hardening commit only — Phase B is a log entry). Apply surviving findings, then proceed.
-### CHECKPOINT 2 — mid-grind (after roughly half the batches) ✅ DONE 2026-07-10
+---
-Full checkpoint protocol completed. Gates green (`make agent-check` + full `make agent-test`, run twice — before and after the review fixes). Cold no-context Sonnet `/code-review` fan-out over `9f5bd54a3..227b98822` (4 agents) ran and was triaged: two real interface/impl-parity defects fixed in `f10b9ad37` (TextFormatRenderable protocol spillover, is_compatible near-symmetric demote), one low-confidence rationale deferred, test changes clean. Record: `wip/subject-grants/checkpoint-2-review.md`. Seeded-remaining count and rubric learnings updated in the Cold-start context above. NEXT: grind cli / cogt / plugins / tools.
+## Stage 0 — inventory + mechanical pre-screen
-### CHECKPOINT 3 — grind complete ✅ REACHED 2026-07-11
+Deterministic checks that need no judgment, to make the semantic pass sharper and cheaper. The seed-ack dogfood showed this class is rich (dead config section, phantom CLI flag, dead config field — all mechanically detectable).
-Zero `seeded = true` entries remain (`--report` confirms `seeded remaining: 0`). Gates green: `make agent-check` + full `make agent-test` green at the checkpoint (and after each of batches 8–11). Batches 8–11 committed (`ea0e642e6` plugins, `c3aa0ee3c` cli, `f542ba941` cogt, `ff8bfbdc6` tools). **CHECKPOINT 3 review DONE + triaged:** cold no-context Sonnet `/code-review` fan-out over `a1649f269..ff8bfbdc6` (3 agents: code-correctness, grant-judgment, mechanical-rewrite-safety). Correctness + mechanical-safety came back clean (independent AST scan = 0 positional-call breakage, all signature rewrites preserved defaults/order/types, no name-collision mis-edits). Grant-judgment found **3 real misses — all fixed**: kept-as-subject lookup-container/scope params where a same-batch sibling demoted the identical shape (`sync_main_config_cmd(target)`, `_resolve_terminal_candidates(deck)`, `suggest_model_alternatives(model_deck)` — the last needed 2 call-site fixes). One low-confidence flag (`get_api_key(backend)` family) considered and deliberately kept (registry-wide "derive value from config" convention). Full record: `wip/subject-grants/checkpoint-3-review.md`. NEXT: Phase 5 finalization.
+- [x] Build the inventory: every in-scope `.md` page (per D2), grouped by section, committed as `wip/drift-hunt/inventory.md` with a page count per section — the denominator for defect density later. **Scope note for Checkpoint 0:** D2's list doesn't name `docs/agents/`, `docs/analytics/`, `docs/distributed-execution/`, nor root-level `changelog.md` / `contributing.md` / `CODE_OF_CONDUCT.md` — inventory them as in-scope-pending-confirmation (hand-written, not generated, not freshly reviewed) and let Louis rule at Checkpoint 0.
+- [x] Write the pre-screen script at `wip/drift-hunt/prescreen.py` (throwaway campaign tooling — self-contained, stdlib + repo venv only). Checks, each producing `(page, claim, verdict, evidence)` rows:
+ - Referenced repo file paths exist in the tree (special-case `docs/cookbook/` → resolve against `../pipelex-cookbook`, per D3).
+ - Mentioned config keys exist in `pipelex/pipelex.toml` / the `configs.py` model tree.
+ - Mentioned `pipelex` / `pipelex-dev` CLI commands and flags exist in the actual `--help` surfaces.
+ - Fenced `toml` blocks parse; fenced `python` blocks compile (`ast.parse`).
+ - Internal doc links resolve (complementing, not re-implementing, the mkdocs strict build).
+- [x] Run it; triage the raw hits into `wip/drift-hunt/suspects.md` — per section: confirmed-mechanical defects vs needs-judgment flags vs false positives (note false-positive patterns; they tune Stage 1 prompts).
+- [x] Fix the confirmed mechanical defects immediately (they need no Stage 1) — one commit, gate: `make docs-check` green.
-## Phase M — Absorb the upstream `dev` merge (⚠ DO THIS BEFORE PHASE 5; then STOP for Louis)
+### CHECKPOINT 0 — STOP (suspect list in hand)
-Louis pushed `feature/Signatures` (tip `c60a448b8`) and is bringing a big update in from `dev`. **Sequencing decision, settled with Louis: merge FIRST, Phase 5 LAST, and check with Louis before starting Phase 5.** Why merge-first: Phase 5 *deletes* the `--seed`/`seeded` machinery, and that machinery is exactly what absorbs the merge's wave of new positional subjects. Close that door only after everyone is through it. Doing Phase 5 first would throw away the onboarding tool at the moment it's most needed AND force redoing the Phase-5 deliverables (CHANGELOG, docs, demoted-surfaces list, drift-ack) that the merge perturbs anyway.
+- [x] Checkpoint protocol steps 1–4: full `make agent-test` green, Stage 0 outputs + fixes committed, Cold-start context updated, `/code-review` fan-out on the fix commit came back clean (all five fixes verified accurate; one adjacent staleness found in the same table — image_reference.py → shared/ — applied as a follow-up commit).
+- [x] **Louis' triage rulings (2026-07-12):** judgment seeds resolved and FP patterns confirmed — recorded as D7/D8, in `suspects.md`, and in `deferred-code-findings.md`.
+- [x] **Louis' scope + shape rulings (2026-07-12):** scope edges all in (D9); Stage 1 = full fan-out, worst-first, in 4 cold-start parts (D10). **Checkpoint 0 CLOSED — each Stage 1 part still needs its own explicit go.**
-### ⚠ The one destructive misstep — fko-clobber, merge edition
+---
-After you resolve the merge, the tree is full of **dev's ungranted positional subjects** — `dev` still runs the OLD generic-exception guard, so every new positional first-arg it added is a violation under our guard. The instant you run **`make fko` or `make agent-check`** (agent-check runs `fko` first), the fixer **silently keyword-only's all of them**, rewriting freshly-merged upstream signatures out from under you. This is the Phase-2 fko-before-seed hazard, now against dev's code.
+## Stage 1 — semantic review fan-out — ✅ **COMPLETE (all 4 parts, 2026-07-12)**
-- **SAFE to run freely:** the read-only gate `make cko` and the PostToolUse edit hook (`.claude/hooks/check-keyword-only.sh`) — both only *report* violations, never fix. Survey with `cko` as much as you like.
-- **SAFE:** `subject-grant --seed` — verified idempotent + additive (`subject_grant_cmd.py::_run_seed`): it seeds only *ungranted* positional subjects and leaves the 1,513 reviewed grants (and any existing seeded entries) untouched; literal-typed subjects are never seeded.
-- **HARD RULE:** run `--seed` **before** the first `fko` / `agent-check` on the merged tree.
+Four parts, one cold-start session each (D10), 104 pages total. Cookbook pages excluded (Stage C, per D7). **Final: 1,439 claims, 109 drift, 3 convention, 0 rejected, density 1.05.** Burn: Part 1 ≈ 4.5M + Part 2 ≈ 4.8M + Part 3 ≈ 2.7M + Part 4 ≈ 2.7M tokens.
-### Merge workflow (cold-start-safe)
+- [x] **Part 1 — building-methods (35 pages). DONE 2026-07-12.** 47 confirmed / 0 rejected → **44 drift** after 3 convention reclassifications (D11). → `findings/building-methods.md`, scoreboard `findings/SUMMARY.md`. Dominant class: doc bundle examples that fail `pipelex validate bundle` even after scaffolding (~18). Strongest drift-contract candidate: `validate-doc-bundle-examples` (with placeholder injection, per D11).
+- [x] **Part 2 — under-the-hood + advanced (26 pages). DONE 2026-07-12.** 42 confirmed / 0 rejected / 0 convention-reclassified → 42 drift (10 breaks / 23 wrong / 4 misleading / 5 stale; density 1.62, or 2.0 vs 21 substantive pages — 5 `advanced/*-injection.md` are Under-Construction stubs). → `findings/under-the-hood-advanced.md`. Dominant class: renamed Python API/method never propagated into docs (the #1 cross-part class now). Strongest new drift-contract candidates: `doc-python-snippets-import` (import+getattr on doc symbols), `doc-python-signatures-vs-live` (keyword-only drift), `doc-toml-config-vs-shipped`.
+- [x] **Part 3 — features + reliability (25 pages). DONE 2026-07-12.** 10 confirmed / 0 rejected / 0 convention-reclassified → 10 drift (2 breaks / 7 wrong / 1 misleading / 0 stale; **density 0.40 — the cleanest section**, 8/25 dirty; all 4 reliability pages clean). → `findings/features-reliability.md`. Dominant class: stale CLOSED enumerations (a list written once, the registry grew, the list never revisited). Strongest new drift-contract candidate: `doc-enumerations-vs-registry`. Two new review rules (exhaustive-vs-illustrative lists; floor counts) — reuse in Part 4. Carry-over → `docs/tools/cli/agent-cli.md:38`.
+- [x] **Part 4 — the tail (18 pages). DONE 2026-07-12.** 12 confirmed by the wave + 1 by the post-run FN check / 0 rejected / 0 reclassified → **13 drift** (2 breaks / 9 wrong / 2 misleading / 0 stale; density 0.72, 9/18 dirty). → `findings/the-tail.md`. Dominant class: **repo-tooling drift** (make targets, generated trees, `mkdocs.yml`, `robots.txt`) — and crucially **not** a phantom-target class: every documented target exists; what drifted is *behavior under a stable name*. Strongest new drift-contract candidate: **`make-targets-and-tooling` — a contract, NOT a derived check** (nothing mechanizable would have caught any of it). Two new rules added (H make-targets, K docs-site meta) plus two scope rules (I process prose out of scope, J version floors). **STAGE 1 COMPLETE ⟶ Checkpoint 1.**
-> **PROGRESS (2026-07-11, second session): steps 0–6 DONE — step 7 REMAINS.** Steps 0–4 ran via the REWIND variant (rewind to pre-grind `9f5bd54a3`, merge dev there — 9 trivial conflicts instead of 23 — then survey/`--seed`/dead-grant cleanup). Step 5 (the regrind) is COMPLETE as waves A/B/C (`fb1a91197` / `65a4153ae` / `f83308367` — see Status line; method in `wip/subject-grants/regrind-after-dev-merge.md`). Step 6 (checkpoint review fan-out + triage) is COMPLETE — findings and triage in the wip record's §Checkpoint review. **NEXT = step 7 (STOP for Louis before Phase 5).**
+Per part (the wave protocol) — **all four parts complete:**
-0. **Decide with Louis:** merge onto `feature/Signatures` directly, or onto a throwaway integration branch off it (keeps the ~1,513-grant baseline trivially recoverable if the merge gets messy). Confirm the exact source ref — Louis said "dev", most likely `origin/dev`; `git fetch` then `git log --oneline origin/dev -5` to eyeball what's coming.
-1. **Merge** (`git merge origin/dev`). Resolve the **mechanism conflicts by keeping OUR side** — this branch *is* the new guard, `dev` has the old one:
+- [x] Per page, a review agent: read the page, extract its checkable claims, verify each against the current code, report findings with severity (`breaks-a-user` / `wrong` / `misleading` / `stale-detail`) and `file:line` evidence for both the doc claim and the contradicting code. Seed each agent with the page's Stage 0 suspect rows and the Checkpoint-0-blessed do-not-flag list (`suspects.md` FP patterns).
+- [x] Adversarial verify pass (D6): every finding independently re-derived by a verifier agent prompted to refute it; only confirmed findings survive. Run as a pipeline (page-review → verify per finding), not a barrier. **Result across Stage 1: 109 confirmed, 0 ever rejected.**
+- [x] **FN spot-check (added in Part 3, promoted to a standing step in Part 4):** hand-verify the *clean* verdicts on the pages seeded as highest-risk — that is where the campaign is most likely to fool itself. It held in Parts 1–3 and **broke a verdict in Part 4** (F13). With 0 rejections across the whole of Stage 1, the pipeline's error mode is **false negatives, not false positives**. Keep this step in Stage C.
+- [x] Aggregate into `wip/drift-hunt/findings/<section>.md` (findings ranked by severity), building toward `findings/SUMMARY.md`: per-section counts, defect density (findings / pages), rejected-findings list with reasons.
+- [x] Log observations relevant to the drift-contracts dogfood — they feed the Phase 3 verdict, per D5. **Headline: drift correlates with *audience*, not complexity** (author-is-the-reader pages stay current; explain-it-to-someone-else pages drift), and **docs edited in the same commit as their code do not drift at all** — which is the drift-contracts thesis, observed in the wild.
+- [x] End of part: commit the findings, update the Cold-start context, report to Louis.
- - `pipelex/cli/dev_cli/commands/keyword_only_guard.py`, `check_keyword_only_cmd.py`, `subject_grant_cmd.py`
- - `pipelex/cli/dev_cli/_dev_cli.py` (the `subject-grant` / `check-keyword-only` wiring)
- - `Makefile` (the `cko` / `fko` / `sgr` / `subject-grant` targets)
- - `docs/contribute/keyword-only-arguments.md` (we rewrote it)
- - `subject_grants.toml` is OURS ALONE (`dev` has no such file) → no textual conflict, but it is now semantically stale wherever `dev` changed code — the guard's symmetric staleness surfaces that in step 2.
- - **Re-splice, do NOT take-ours wholesale:** `.claude/rules/python-standards.md`, this repo's `CLAUDE.md`, and `drift.toml` — take **dev's** version of each file (it carries dev's other evolution, e.g. Python-3.10-drop wording and any new drift contracts), then re-splice ONLY our keyword-only section / `keyword-only-convention` contract back in. Do not clobber dev's non-keyword content.
+### CHECKPOINT 1 — ✅ CLOSED (Louis, 2026-07-12)
-2. **Survey (read-only, no clobber):** `make cko`. Expect four violation kinds — `ungranted-subject` (dev's NEW positional subjects — the big wave), `dead-grant` (a grant whose def dev renamed/moved/deleted), `grant-param-mismatch` (dev changed a granted def's first param name), `literal-subject` (dev added a `bool`/`int`/`float` subject).
-3. **Onboard the new wave:** `.venv/bin/pipelex-dev subject-grant --seed` (additive — safe). Stamps every new ungranted positional subject as `seeded = true`. `--report` then shows per-package seeded counts — the Phase-4 progress meter, now for the merge delta.
-4. **Fix the staleness `--seed` does NOT handle:** `dead-grant` → delete the entry (def gone) or re-grant at the new path (def moved but still a legit subject); `grant-param-mismatch` → update the entry's `param` if the subject was just renamed, else demote; `literal-subject` → demote the def (or `# kw-only: ignore` if framework-positional) — never seed it (D4).
-5. **Now** `make agent-check` is safe (seeded entries are valid grants; `fko` leaves them alone). **Grind the new seeded wave batch-by-batch** with the exact Phase-4 **Batch execution recipe** + this session's mechanical tooling (signature-rewriter + call-site fixer — see batch-8–11 learnings item (g)), applying the same rubric + all accumulated case-law.
-6. **Gates + checkpoint review:** `make agent-check` + full `make agent-test` green; fan out a fresh no-context `/code-review` over the merge+grind commit range (same 3-agent shape as CHECKPOINT 3: code-correctness, grant-judgment, mechanical-rewrite-safety); triage real defects, record under `wip/subject-grants/`.
-7. **⚠ STOP — confirm with Louis before starting Phase 5.** Do NOT auto-proceed. Phase 5 closes the transitional door (deletes `--seed`); only do it once Louis agrees the merge is fully absorbed and the tree is clean (`--report`: `seeded remaining: 0` again, all gates green).
+- [x] Checkpoint protocol steps 1–4 (findings committed `32792618d`; full `make agent-test` green; review fan-out n/a — no code/tooling changed).
+- [x] **Louis' rulings — recorded as D12–D17 below.** Stage C deferred; the `agent-test` heartbeat gets fixed everywhere (incl. both `CLAUDE.md`s) in Stage 2; Stage 2 = 4 parts, one PR each; fixers may restructure a broken example; mid-fix drift is fixed in place on the same page; Stage 2 stays docs-only.
-## Phase 5 — Finalization
+---
-**⚠ GATED: do NOT start until Phase M (the `dev` merge) is fully absorbed AND Louis has signed off.** Phase 5 deletes `--seed`, which the merge depends on — see Phase M.
+## Stage C — cookbook cross-check — ⏸️ **DEFERRED (Louis, 2026-07-12 — D12)**
-- [x] Delete `--seed` and all `seeded`-field support; tighten the registry schema (exactly `param` + `rationale`; unknown keys fail the check). The now-dead `collect_all_def_infos` helper deleted too (the dead-grant scan caught its stale entry — the symmetry working as designed).
-- [x] CHANGELOG `[Unreleased]`: convention hardened (grants registry, literal-subject ban) + breaking note covering the demoted signatures.
-- [x] Final docs pass (convention doc reflects the post-transitional state; no `seeded` mentions survive); drift re-acked (keyword-only-convention + cli-docs + config-docs, the last two opened by the regrind / dev merge).
-- [x] Hand off the demoted-public-surfaces list to the release-gated cross-repo sweep: `wip/subject-grants/demoted-public-surfaces.md`.
-- [x] Full gates: `make check` end-to-end + `make agent-test`. Cold-start context updated to COMPLETE.
+Not part of this campaign's remaining arc. The 21 `docs/cookbook/` pages stay **unswept**; note that in the close-out so the campaign's coverage claim stays honest (Stage 1 covered 104 of the 125 in-scope pages). If it is ever revived: compare each page against `../pipelex-cookbook` (read-only ground truth, D3), reuse the Part-4 wave script (it carries the full rule stack), and budget the FN spot-check as a first-class step.
-## Risks & safety nets
+---
-- **Framework-positional callers** (Typer `callback=`, Jinja filters, SDK hooks): pyright passes a wrongly-kwonly'd callback; `make agent-test` is the net — mandatory after the Phase 2 sweep and any suspicious Phase 4 demotion.
-- **fko-before-seed hazard** (Phase 2 warning above) — the one truly destructive misstep; land guard + seed atomically. **Recurs at the `dev` merge (Phase M):** after resolving the merge, `--seed` the new ungranted subjects BEFORE the first `fko`/`agent-check`, or the fixer silently keyword-only's all of dev's freshly-merged positional signatures. `make cko` (read-only) and the edit hook are safe; `--seed` is additive/idempotent.
-- **Rubber-stamp risk in the grind:** rationale honesty can't be mechanized. Mitigations: def-specific rationale rule, Louis spot-checks registry diffs, batches small enough to review for real.
-- **Registry merge conflicts** across parallel branches: sorted, append-mostly file; conflicts resolve trivially, same as the drift manifest.
-- **Hook cold-start budget:** one `tomllib` read of a machine-written file — negligible; keep the lean path import-clean (no tomlkit there).
+## Stage 2 — fix in batches (shape settled by D14; **4 parts, one cold-start session each**)
-## Out of scope
+Mirrors Stage 1's session-sharding: each part resumes cold from **its own findings file alone**, lands as **one PR to `dev`** (D14), and touches a page set disjoint from the other parts — so the four can land in any order without conflicts.
-- Any LLM verdict inside CI (D3 — permanent).
-- Other repos' conventions; the conformance suite (this is repo-internal dev tooling, no cross-repo spec surface).
-- Retro-fitting the symmetric-tuple allowlist into the registry (D8 — different semantics, stays in code).
+**Recommended order: S2-3 → S2-1 → S2-2 → S2-4.** S2-3 first because it is the highest volume but the *lowest effort per fix* (mostly one-line renamed-API substitutions) — it shakes out the fix loop cheaply. S2-1 second, while attention is freshest, because it is the hardest. S2-4 last because it absorbs every cross-cutting sweep.
+
+- [x] **S2-1 — building-methods / `pipes/`** (30 findings, 17 pages). **DONE 2026-07-12 — PR #1043 MERGED to dev (squash `126616a88`)** (branch `docs/Drift-hunt-s2-1`, deleted; campaign-branch commits `7942339cc` + `dd6174ee6`). All 30 fixed, 0 rejected-at-fix-time; every changed bundle example re-validated against the real parser (incl. dry-run), broken examples restructured per D15 (PipeSequence, PipeBatch, PipeParallel combine, PipeCondition, multiplicity, pipes/index). +F45–F54 fixed per D16 (11 candidates → 10 confirmed, 2 refuted). Blind post-fix review caught 2 defects in the batch's own text (the non-running `bundle_uris` run-a-bundle snippets on two pages; the F45 dedup gate) — fixed, snippets executed live in dry-run. 2 code-side deferrals recorded (items 6–7). The PR carries this page set's previously-unshipped Stage-0 fix, self-contained on dev. Full fix-time record in `findings/building-methods.md`.
+- [x] **S2-2 — building-methods / concepts + top-level** (14 findings, 9 pages). **DONE 2026-07-12 — PR #1044 MERGED to dev (squash `46511ce9c`)** (branch `docs/Drift-hunt-s2-2`, deleted; campaign-branch commits `8f03eb6f2` + `1824f2db4`). All 14 fixed, 0 rejected-at-fix-time; every changed example re-validated against the real parser (libraries.md `execute()` + F3 sequence live in dry-run). +F55–F61 fixed per D16 (12 candidates → 7 confirmed + 1 out-of-definition stale pointer, 6 refuted). Blind post-fix review: zero defects. 2 code-side deferrals (items 8–9); 1 out-of-batch defect recorded (PipeLLM.md inline `@document`, S2-1 page). The PR carries this page set's previously-unshipped Stage-0 fixes (incl. `adapt-to-llm-prompting-style-*.md`), self-contained on dev. Full fix-time record in `findings/building-methods.md`.
+- [x] **S2-3 — under-the-hood + advanced** (42 findings, 26 pages). **DONE 2026-07-12 — PR #1042 MERGED to dev (squash `9814e2f4c`)** (branch `docs/Drift-hunt-s2-3`, deleted; campaign-branch commits `10b4bae0a` + `af7106342`). All 42 fixed, 0 rejected-at-fix-time; +F43–F46 fixed per D16 (2 more candidates refuted — first refutations of the campaign); F24 handled per the Cold-start note; 2 code-side deferrals recorded. The PR also carries these pages' previously-unshipped Stage-0 fixes, keeping it self-contained on dev. Full fix-time record in `findings/under-the-hood-advanced.md`. **Post-PR bot-review triage (2026-07-12):** greptile+cubic left 6 unresolved threads → deduped to 4 issues, all 4 confirmed against the code (2 incomplete config tables now presented-as-exhaustive, 1 table row contradicting the corrected `--graph-no-data` info block, 1 "copied once" vs full-reset wording), fixed in PR-branch commit `532725251`, mirrored here via cherry-pick; all 6 threads replied + resolved.
+- [x] **S2-4 — features + reliability + the tail + ALL cross-cutting sweeps** (23 findings, 25 pages + the sweeps). **DONE 2026-07-12 — PR #1045 MERGED to dev (squash `cb13e5b33`)** (branch `docs/Drift-hunt-s2-4`, deleted; finalization commit `1b776eca5` = Louis's review edits + 8 bot threads triaged; campaign-branch mirror `5e15bec08`). All 23 fixed, 0 rejected-at-fix-time; all four cross-cutting items landed (PipeStructure roster ×3, `agent-cli.md:38`, D13 heartbeat everywhere — incl. the `pipelex/kit/agent_rules/` SOURCE templates + `make rules` regeneration of `CLAUDE.md`/`AGENTS.md` after `check-rules` caught the direct edit, and the workspace-root `CLAUDE.md` in its own repo, commit `cc2b68c` on `feature/Follow-ups`; `docs/setup/telemetry.md` untouched). +F11 + F14–F22 fixed per D16 (13 candidates → 10 confirmed, 3 refuted). Blind post-fix review: zero defects; one snippet-import defect in the batch's own text caught pre-commit. 1 code-side deferral (item 10); 1 out-of-batch record (observer-provider-injection.md + advanced/index.md, S2-3 pages). Full fix-time record split across `findings/features-reliability.md` and `findings/the-tail.md`.
+
+Per batch:
+
+- [ ] Fix highest severity first within the part. Every fix **re-verified against the code at fix time** (D6) — the finding is the pointer, **the code is the authority**. A finding that no longer reproduces is marked rejected-at-fix-time, not silently patched.
+- [ ] Gates: `make docs-check` (mkdocs strict) + `make check`. ⚠ **`docs-check` does NOT cover `CONTRIBUTING.md`, `CODE_OF_CONDUCT.md` or the `CLAUDE.md`s** — they are not mkdocs pages. Verify those by hand.
+- [ ] Fixing docs does NOT reopen drift contracts (docs are review targets, not triggers). If a fix touches a trigger file, resolve via the `drift-review` skill.
+- [ ] Mark resolved findings in the findings files as batches land (mark, don't delete — audit trail until close-out).
+- [ ] Per-batch `/code-review` fan-out (checkpoint protocol step 4) before the PR is finalized.
+- [ ] Open the PR to `dev` (D14), then report to Louis. Session ends; the next part runs cold.
+
+### CHECKPOINT 2 — ✅ EXECUTED (2026-07-12) — STOPPED for Louis' final review
+
+- [x] All batches landed (all four PRs merged to dev 2026-07-12, campaign branch reconciled, `make docs-check` + `make check` green on the reconciled branch) — **full `make agent-test` on this branch: GREEN** (and the run displayed the start-line + heartbeat lines the D13 fixes document).
+- [x] Close-out written in `findings/SUMMARY.md`: coverage disclosure (**104 of 125 in-scope pages swept; the 21 cookbook pages were NEVER swept, D12**), Stage-2 resolution table (109/109 fixed, 0 rejected-at-fix-time; +31 fix-time findings + 1 stale pointer fixed, 13 refuted), the ranked drift-contract shortlist handed to the Phase 3 verdict (per D5/D17), and the campaign's lessons.
+- [x] `wip/drift-hunt/deferred-code-findings.md` (10 items) handed off — kept in place, header updated. **The "partial out-of-batch observer page" turned out to be a stale-base artifact, NOT a real remainder:** verified at close-out, `docs/advanced/observer-provider-injection.md` (and `advanced/index.md`) is fully fixed and byte-identical to dev — Part 2's F38–F42 covered those exact defects and S2-3 (PR #1042, squash `9814e2f4c`) fixed them; S2-4's session recorded them as "remaining" because its base predated #1042's merge. Verified against live code: hooks `observer_protocol.py:17-31`, payload keys `pipe_router_protocol.py:18-47`, JSONL names `local_observer.py:11-14`, import + `Pipelex.make` kwargs + `hub.set_observer` (`hub.py:263`). No follow-up needed.
+- [x] `wip/drift-hunt/` pruned: kept `findings/SUMMARY.md` + `inventory.md` (the record) + `deferred-code-findings.md` (the handoff); removed the four per-section findings files, `prescreen.py`, `prescreen-raw.md`, `suspects.md`, `checkpoint-0-decision-brief.md` — all recoverable at `git show drift-hunt-archive:wip/drift-hunt/<path>` (pushed tag pinning `9a6201a5e`).
+- [x] Cold-start context updated to DONE.
+- [ ] **STOP for Louis:** final review of the close-out; decide what happens to the campaign branch (do NOT push unless asked).
+
+---
+
+## Out of scope (explicit)
+
+- cocode, in any role (D1).
+- Growing `drift.toml` (D5 — evidence now, contracts later).
+- Editorial/style/structure rewrites, new content, SEO.
+- Generated docs (`docs/errors/`, gateway models) — fix the generator if one of those is wrong.
+- `docs/configuration/` + CLI docs (just reviewed via seed acks) — unless a Stage 1 finding elsewhere points back into them.
+- Other repos' docs (read `../pipelex-cookbook` as ground truth; don't fix it — hand defects off via SUMMARY.md).
diff --git a/docs/contribute/drift-contracts.md b/docs/contribute/drift-contracts.md
index 995dfe456..f8b2cb943 100644
--- a/docs/contribute/drift-contracts.md
+++ b/docs/contribute/drift-contracts.md
@@ -28,6 +28,20 @@ The validity rule is a single equality, with no base ref, no git diff, and no ti
The digest covers both the contract's own definition and the `(path, blob OID)` pairs of every tracked file matching its triggers — so editing a trigger file, adding or deleting a matched file, *or editing the contract itself* all open the contract. A brand-new contract has no ack, so it is open until its targets are reviewed once and acked: adoption is explicit, contract by contract.
+## Where the gate acts
+
+One view of every surface the contract system touches, from the working tree to the merge:
+
+| Surface | What happens there |
+|---|---|
+| **Make targets** | `make drift-check` (alias `dc`) is the pass/fail gate; `make drift-plan` (`dp`) is the read-only diagnosis; `make drift-ack CONTRACT=… RATIONALE="…"` (`da`, optional `BY=` for agents) records the review. All three wrap `pipelex-dev drift`. |
+| **Quality checks** | `drift-check` runs inside the `make check` aggregate, alongside the other repo gates — a full local check cannot pass with an open contract or a rotten manifest. |
+| **Commits** | The digest is computed from the **git index**, and `drift ack` stages the ack file it writes — so the ack, the code change, and any doc fix land in the same commit, reviewed as one diff. |
+| **Branches / merges** | After a merge, `drift check` recomputes the digest over the merged tree. An ack that was valid on either side but does not cover the merged trigger content fails the check until the merged state is reviewed and re-acked (see [Merges](#merges)). |
+| **PRs / CI** | The `lint-drift` CI job runs `make drift-check` on every PR — currently advisory (visible, not merge-blocking; promotion to a required check is planned). The committed ack file appears in the PR diff, so the reviewer sees the rationale next to the change it covers. |
+
+The gate acts on *content*, never on *time*: there is no base ref, no diff against a branch, no timestamp freshness — only the digest equality below.
+
## The ack workflow
When `make drift-check` fails (locally or in CI), run:
diff --git a/mkdocs.yml b/mkdocs.yml
index 1ffcbe783..3c075bee2 100644
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -545,6 +545,7 @@ nav:
- Overview: under-the-hood/index.md
- Architecture Overview: under-the-hood/architecture-overview.md
- Build-time Elaboration: under-the-hood/build-time-elaboration.md
+ - Codegen Projections: under-the-hood/codegen-projections.md
- Dry Run Mock Generation: under-the-hood/dry-run-mock-generation.md
- Execution Graph Tracing: under-the-hood/execution-graph-tracing.md
- Image Handling in LLM Prompts: under-the-hood/image-handling-in-llm-prompts.md
diff --git a/wip/drift-contracts/README.md b/wip/drift-contracts/README.md
new file mode 100644
index 000000000..2ca832ada
--- /dev/null
+++ b/wip/drift-contracts/README.md
@@ -0,0 +1,8 @@
+# Drift contracts — track folder
+
+Working notes for the drift-contracts system (the shipped system itself is documented at `docs/contribute/drift-contracts.md`; the original design rationale at `../drift-contracts-design.md`).
+
+- **`phase-3-verdict.md`** — the pilot verdict working doc: per-contract keep/narrow/mechanize/drop rulings, trigger repairs, new contracts, derived-check adoption. **The live document — awaiting Louis' rulings.**
+- **`dogfood-log.md`** — one mandatory entry per `drift ack` (real-catch / clean-pass / friction). The pilot's primary evidence stream; stays mandatory per the verdict doc's E3.
+
+Campaign-side evidence (the Drift Hunt) lives in `../drift-hunt/findings/SUMMARY.md` — its Handoff 1 is the drift-contract shortlist the verdict doc consumes.
diff --git a/wip/drift-contracts/phase-3-verdict.md b/wip/drift-contracts/phase-3-verdict.md
new file mode 100644
index 000000000..3b9dbed12
--- /dev/null
+++ b/wip/drift-contracts/phase-3-verdict.md
@@ -0,0 +1,176 @@
+# Drift contracts — Phase 3 pilot verdict (working doc)
+
+**Status: DRAFT — awaiting Louis' rulings (decision boxes below).** This is the deferred "Phase 3 — pilot verdict" item from the Drift Hunt tracker (`TODOS.md` → Deferred drift-contracts items): the per-contract **keep / narrow / mechanize / drop** decision the design said must wait for dogfood evidence, plus the manifest-growth question that evidence now answers. Ruling on this doc formally ends the D5 dogfood freeze ("evidence now, contracts later" — the campaign is closed, the evidence is in).
+
+Naming note: the design doc's rollout plan (`wip/drift-contracts-design.md` → Rollout plan) used "Phase 3" for *agent ergonomics, after the dogfooding verdict*; the campaign tracker uses "Phase 3" for the verdict itself. This doc is the verdict; the agent-ergonomics leftovers are folded in as agenda item E3. (The tracker's pointer to the original phase breakdown — `git show origin/docs/Update:TODOS.md` — is dead, that branch is gone; the surviving source is the design doc's Rollout plan section.)
+
+## The question the pilot asked
+
+From the design doc, verbatim: *"is the ack friction proportionate to the staleness it catches?"* — with the promised consequence: *"If the dogfood shows the acks catching real staleness at tolerable friction, grow the manifest; if it shows ceremony, we will have learned that cheaply."*
+
+**Verdict: the mechanism works; grow the manifest — narrowly.** The evidence below supports keeping all three seed contracts, repairing identified trigger gaps, adding two new contracts, and adopting the campaign's top derived checks. It does not support broad trigger expansion (see B5).
+
+## A — Evidence base
+
+Three independent sources:
+
+**A1 — the dogfood log** (`wip/drift-contracts/dogfood-log.md`). Every ack so far logged: one **real-catch** (keyword-only — the guard gained the subject-grant registry and the convention doc was materially stale, exactly the coupling the contract exists to enforce), the rest **clean-pass** (genuine reviews, nothing stale), and **zero friction verdicts** (no contract ever opened on changes that could not have affected its targets). Small sample, but the directional read is clean: real staleness caught, no ceremony observed. One process slip (an ack recorded without its mandatory log entry, backfilled honestly) — an argument for keeping the log mandatory, not for dropping it (see E1).
+
+**A2 — the Drift Hunt campaign** (`wip/drift-hunt/findings/SUMMARY.md`, close-out). The campaign is the system's external validation:
+
+- **Docs edited in the same commit as their code did not drift** — observed in the wild (`d3550330a`, `15e0a94ea`). That is the drift-contracts thesis, confirmed empirically.
+- **Drift follows audience, not complexity.** Author-is-the-reader pages stay current by use; explain-it-to-someone-else pages drift at up to 2.0 findings/page. Contract budget belongs on reader-facing pages.
+- **Some drift classes are provably un-mechanizable.** Part 4's repo-tooling class: every documented make target exists (an existence script scores zero), yet behavior drifted under stable names — `agent-test` grew a heartbeat weeks after a page was written on the premise it was silent. The right instrument is a review obligation. This is the campaign's strongest evidence *for* the contracts mechanism itself.
+- **A green snippet check must never be read as a clean page** (F13: four bundles validate and dry-run clean while the prose beside them teaches a false rule). This bounds what the derived checks in section D can be trusted to mean.
+- The ranked drift-contract shortlist handed to this verdict (SUMMARY.md → Handoff 1).
+
+**A3 — the trigger-coverage audit** (this session, 2026-07-13). Cross-checking `drift.toml`'s trigger globs against where the documented surfaces actually live in the code found the concrete gaps itemized in section B.
+
+## B — Verdicts on the seed contracts
+
+| Contract | Verdict | Change |
+|---|---|---|
+| `config-docs` | **KEEP + repair triggers** | Add the inference-backend/routing surface (B1); adopt the coverage meta-check (section C) |
+| `cli-docs` | **KEEP as-is** | None — right glob, right exclusion, right targets |
+| `keyword-only-convention` | **KEEP + extend trigger** | Add `subject_grant_cmd.py` (B3) |
+
+### B1 — `config-docs`: repair the trigger enumeration
+
+**The hole today:** `docs/configuration/config-technical/inference-backend-config.md` documents the backend TOML layout, per-backend setup, gateway overrides, and routing profiles — whose sources of truth are `pipelex/cogt/model_backends/*.py`, `pipelex/cogt/model_routing/*.py`, and the shipped structural TOMLs under `pipelex/kit/configs/inference/`. None of those are triggers; only `config_cogt.py` is. A field added to the `Backend` model, or a restructuring of `backends.toml` / `routing_profiles.toml`, changes what that page must say without opening the contract.
+
+**Proposed trigger additions:**
+
+```toml
+"pipelex/cogt/model_backends/**/*.py",
+"pipelex/cogt/model_routing/**/*.py",
+"pipelex/kit/configs/inference/backends.toml",
+"pipelex/kit/configs/inference/routing_profiles.toml",
+```
+
+**Deliberately excluded:** the per-backend deck files (`pipelex/kit/configs/inference/backends/*.toml`). The add-model flow touches those routinely; making every model addition cost an ack is exactly the friction the manifest header warns against, and the docs page documents the *structure*, not the model roster. Residual risk accepted: a backend file rename could drift the doc's examples without opening the contract — the enumerations derived check (D1) is the better instrument for roster-shaped drift.
+
+**The structural fragility:** hand-enumerating submodel files means every *new* documented submodel defined in its own package silently escapes the gate — drift of the drift manifest, the one place the system cannot see itself. The settings-model files (`llm_setting.py`, `extract_setting.py`, `img_gen_setting.py`, …) are the current long tail: some are documented via sections that live in already-triggered files, some may not be. Rather than audit them by hand once (and rot again), fix it structurally — see section C.
+
+- [ ] **DECISION B1 (Louis):** adopt the four trigger additions? Include or exclude the per-backend deck files?
+
+### B2 — `cli-docs`: keep as-is
+
+The glob (`pipelex/cli/**/*.py` minus `dev_cli/`), the review targets, and the observed behavior (a clean-pass ack on a broad internal sweep that changed no command surface) are all as designed. No change proposed.
+
+- [ ] **DECISION B2 (Louis):** confirm keep-as-is.
+
+### B3 — `keyword-only-convention`: the trigger is one file, the convention's tooling is two
+
+The doc specifies the grant workflow (`make sgr`), implemented in `pipelex/cli/dev_cli/commands/subject_grant_cmd.py` — not in the guard. Not hypothetical: the Signatures track deleted `--seed`/`seeded` from exactly that command and the doc had to follow (it did, by discipline).
+
+**Proposed:** add `"pipelex/cli/dev_cli/commands/subject_grant_cmd.py"` to the contract's triggers.
+
+- [ ] **DECISION B3 (Louis):** adopt.
+
+### B4 — a note on the keyword-only contract's audience
+
+The campaign's headline correlation says maintainer-facing pages stay current by use, and `keyword-only-arguments.md` is maintainer-facing — it came back clean in Stage 1. That cleanliness is confounded (the contract already existed and had produced the pilot's one real-catch), and the contract's friction is near zero (a single-file trigger that rarely changes). **Proposed: keep.** Recorded here so the "narrow or drop" question was asked on the record, not skipped.
+
+### B5 — what NOT to do: no broad Python-surface triggers
+
+The worst drift density was in the under-the-hood docs (renamed Python APIs, ~2.0 findings/page), and the tempting response is a contract triggered on the API surface. The trigger would be effectively `pipelex/**/*.py`, it would open on every PR, and acks would degrade into rubber stamps — the exact failure mode the design doc names ("if we observe reflexive acks, the contract is mis-scoped"). That drift class belongs to the derived checks (section D), which is also where the campaign's shortlist puts the top value-per-effort. **Proposed: no contract for the Python API docs; adopt D3 instead.**
+
+- [ ] **DECISION B5 (Louis):** confirm.
+
+## C — New: manifest-coverage meta-check (keeps B1 from rotting)
+
+**Proposed:** a unit test (not a new CLI command — smallest correct surface) that walks the tree for files defining `ConfigModel` subclasses, matches each against `config-docs`' trigger globs, and fails on any file that is neither matched nor on an explicit not-user-documented allowlist (factories, internal specs, exceptions). Lives with the drift tests, runs in `make agent-test`, needs no Makefile or CI wiring. When a new documented submodel appears in its own package, the test fails until either the trigger list or the allowlist is updated — the enumeration can no longer rot silently.
+
+The allowlist seeding is the one judgment call: it requires classifying today's `ConfigModel` files as documented-vs-internal once, which is also the audit that settles B1's long tail (the settings-model files).
+
+- [ ] **DECISION C (Louis):** adopt the meta-check as a unit test?
+
+## D — Derived checks to adopt (from the campaign shortlist)
+
+Per the manifest's own rule, anything mechanizable becomes a derived check, never a contract. The shortlist, with a proposed build order (by effort, cheapest first — the shortlist's own ranking is by value-per-effort and puts D1 first; D3 is cheaper to build, so the two orderings differ only in which ships first):
+
+**D3 — `doc-python-snippets-import`** *(build first — lowest effort)*: import-and-`getattr` smoke check on Python symbols referenced in doc snippets. Kills the campaign's #1 cross-part defect class (renamed API never propagated into docs), which produced the worst silent failure found anywhere (mis-named observer hooks that silently no-op).
+
+**D1 — `doc-enumerations-vs-registry`** *(build second)*: diff documented **closed** lists against the code's registries (native concepts, operator roster, model deck, config files). The diff is trivial; the design cost is encoding the exhaustive-vs-illustrative rule — only hedge-free lists are checked ("such as" / "including" / a scope qualifier exempts a list; two S2-4 refutations prove skipping scoped lists is correct, not lenient).
+
+**D2 — `validate-doc-bundle-examples`** *(build third — highest impact, most engineering)*: extract fenced bundle examples under `docs/`, inject D11 placeholders (`description`/`domain`, tolerated-omitted prompt), scaffold D19 placeholder sub-pipes for sequence sketches, run `pipelex validate bundle`. Kills the single biggest breaks-a-user class. **Carry the F13 caveat into its output and docs: a green run verifies the snippets, never the sentences beside them.**
+
+**Deferred (weaker / harder, revisit only if their classes recur):** `doc-python-signatures-vs-live` (keyword-only drift in snippets; needs signature resolution), `doc-toml-config-vs-shipped` (diff doc-embedded `pipelex.toml` blocks against the shipped file), `operator-param-table-vs-blueprint` (needs a Markdown table parser).
+
+Wiring, for each adopted check: a `pipelex-dev` command mirroring the existing `check_*` family, a Make target, membership in the `make check` aggregate (hence CI). Each ships as its own PR with the docs-tree fixes its first run demands (expected to be near-zero — Stage 2 just fixed everything these checks would catch, which also makes now the cheapest moment to adopt them: they start green and stay green).
+
+- [ ] **DECISION D (Louis):** adopt D3 + D1 + D2 in that order? All three, or a subset?
+
+## E — New contracts
+
+### E1 — `make-targets-and-tooling` (shortlist #4 — the clearest contract case the campaign produced)
+
+The un-mechanizable class from A2: behavior drifting under stable target names. **Proposed contract:**
+
+```toml
+[contracts.make-targets-and-tooling]
+description = "Pages that describe make targets must track the Makefile — behavior drifts under stable target names."
+triggers = ["Makefile"]
+review = [
+ "docs/agents/debugging-hanging-pytest-runs.md",
+ "docs/contribute/drift-contracts.md",
+ "docs/contribute/keyword-only-arguments.md",
+ "CONTRIBUTING.md",
+ "pipelex/kit/agent_rules/commands.md",
+ "pipelex/kit/agent_rules/codex_commands.md",
+]
+verify_commands = []
+```
+
+Notes: the kit agent-rules templates are the *source* of the generated repo-root `CLAUDE.md`/`AGENTS.md` — review the templates, `make rules` regenerates (the S2-4 lesson). Exact review-target list to be settled at implementation by grepping for pages that *describe* targets rather than merely invoke one (the seed contracts set this precedent). Review-target overlap with other contracts (e.g. `keyword-only-arguments.md` under both this and `keyword-only-convention`) is fine — contracts are independent obligations.
+
+**Friction, stated honestly:** the Makefile changed 13 times in the last three months (6 in the last month) — this contract opens roughly weekly in active periods. Most acks will be the one-minute "changed recipe isn't documented" kind. If the dogfood shows reflexive acks, the design doc's remedy applies: narrow (e.g. trigger only on documented target names' recipe blocks is *not* mechanically expressible — so realistically: shrink the review list, or accept and re-verdict).
+
+- [ ] **DECISION E1 (Louis):** adopt? Confirm or trim the review-target list.
+
+### E2 — `drift-tool-docs` (self-hosting)
+
+`dev_cli` is excluded from `cli-docs` (correctly), and the keyword-only guard got its own carve-out contract — but the drift tool itself didn't, though `docs/contribute/drift-contracts.md` + the drift-review skill are the same "the doc is the tool's specification" shape. Phase H changed ack semantics (auto-stage, warn→fail) and updated both docs in the same change — by discipline, not obligation. The tool that enforces review obligations should carry one for its own spec. **Proposed contract:**
+
+```toml
+[contracts.drift-tool-docs]
+description = "The drift-contracts doc and the drift-review skill are the tool's specification and must track its behavior."
+triggers = ["pipelex/cli/dev_cli/commands/drift/**/*.py"]
+review = [
+ "docs/contribute/drift-contracts.md",
+ ".claude/skills/drift-review/SKILL.md",
+]
+verify_commands = []
+```
+
+Friction: near zero (the drift subpackage changed twice in the last month, both in one hardening wave).
+
+- [ ] **DECISION E2 (Louis):** adopt.
+
+### E3 — standing policy items
+
+- **Dogfood log:** keep the mandatory one-line entry per ack through the next wave — the repaired triggers and new contracts need their own evidence, and the E1 friction estimate needs confirming in practice. Revisit once the new contracts have accumulated enough acks to judge (suggested trigger for the revisit: the first friction verdict, or a quarter, whichever comes first).
+- **`lint-drift` promotion to a required CI check:** stays gated (tracked in `TODOS.md`). Flip condition: every open PR against `dev` either merged or rebased past the current acks — a required check would fail every PR that predates them. Re-check the backlog after the manifest changes from this verdict land, since those add new acks and reset the clock.
+- **Agent ergonomics (the design doc's original "Phase 3"):** substantially delivered — the drift-review skill exists, the kit rules describe the workflow, and the failure messages are actionable. The remaining design-doc item (`--format json` for software consumers) is deferred until a software consumer exists (clean-solid rule: no speculative surface).
+
+- [ ] **DECISION E3 (Louis):** confirm the three policy positions.
+
+## F — Execution plan (once rulings are in)
+
+**Phase V1 — manifest wave (one PR).** Apply B1 + B3 trigger repairs and add the E1 + E2 contracts to `drift.toml`; perform the initial reviews *for real* (the Phase-2-seeds precedent — note the make-targets pages were just swept and fixed by the campaign, so the first reviews should be fast but must still be genuine); record the initial acks + dogfood-log entries; update `docs/contribute/drift-contracts.md`'s contract roster and the changelog.
+
+**Phase V2 — meta-check (same PR or a small follow-up).** The section-C unit test, including the one-time documented-vs-internal classification of `ConfigModel` files that seeds its allowlist.
+
+### CHECKPOINT V-A
+
+Gates green (`make agent-check`, targeted drift tests, `make docs-check`); PR to `dev`; update this doc's status and the memory. Natural handoff — V3 is independent engineering a fresh session can pick up cold from section D.
+
+**Phase V3 — derived checks (one PR each, in D's build order).** Each: `pipelex-dev` command + Make target + `make check` membership + docs; first run against the docs tree, fix or triage anything it flags.
+
+### CHECKPOINT V-B
+
+All adopted checks green in CI. Re-check the `lint-drift` promotion gate (E3) against the then-current PR backlog; if clear, promote in its own small PR.
+
+## Bottom line
+
+The pilot answered its question: acks catch real staleness (one real-catch, zero friction verdicts, zero observed rubber-stamping), and the campaign independently confirmed both the thesis (same-commit docs don't drift) and the existence of drift classes only a review obligation can cover. The proposed growth is deliberately narrow — two contracts, four trigger repairs, three derived checks, one meta-check — and everything broad stays out, per the design's own friction principle.
diff --git a/wip/drift-contracts/pr-1040-review-notes.md b/wip/drift-contracts/pr-1040-review-notes.md
deleted file mode 100644
index 3ecbf67ac..000000000
--- a/wip/drift-contracts/pr-1040-review-notes.md
+++ /dev/null
@@ -1,25 +0,0 @@
-# PR #1040 review notes — deferred drift-tooling findings
-
-> **RESOLVED 2026-07-12** — both findings were implemented on `docs/Drift-hunt` (commit `8c7413afc`): A1's auto-staging verbatim, A2 as the narrow warn→fail escalation recorded below (not codex's blanket fail). The "why deferred" rationales underneath are the historical record of the original triage, no longer current.
-
-Two codex findings on `pipelex/cli/dev_cli/commands/drift/drift_cmd.py` from the PR #1040 review pass. Both describe real code behavior (verified against the code), but neither is worth hardening now: the drift tool is an internal dev CLI, CI backstops the first, and the second is a documented design choice with a narrow failure mode. Captured here so the candidate fixes aren't lost.
-
-## A1 — `drift ack` does not stage the ack file, so local `drift check` can false-green
-
-Reporter: codex (P2). Thread: <https://github.com/Pipelex/pipelex/pull/1040#discussion_r3564484373>
-
-`drift check` hashes trigger files from the **git index** but loads ack files from the **working tree** (`load_all_acks` globs `.drift/acks/*.toml` on disk). `drift ack` writes the ack file without staging it. So right after an ack, the local gate passes even though the ack blob is not in the commit being built — if the developer forgets to `git add` the ack, the false green only flips red in CI (fresh checkout: `MISSING_ACK` / `DIGEST_MISMATCH`).
-
-**Why deferred:** the false green is purely local; CI deterministically catches anything that lands. The doc already instructs committing the ack together with the change. Dev-loop papercut, not a correctness hole.
-
-**Recorded candidate fix (clean, low-risk):** have `drift ack` auto-stage the ack file it writes (`git add .drift/acks/<id>.toml` after `save_ack`). The ack is a tool-written artifact, never hand-edited, so auto-staging is safe; it makes the local gate honest (the ack lands in the same index the check reads) and removes the forgot-to-add footgun. Note: the alternative "read acks from the index" would report `MISSING_ACK` immediately after a successful ack unless combined with auto-staging, so auto-staging is the load-bearing part. Test approach: temp-git-repo test asserting the ack path appears in `git diff --cached --name-only` after `drift ack`.
-
-## A2 — verify commands run against the working tree while the ack digest covers staged blobs
-
-Reporter: codex (P2). Thread: <https://github.com/Pipelex/pipelex/pull/1040#discussion_r3564484375>
-
-In `drift_ack_cmd`, verify commands run against the working tree, then the digest is computed from staged blob OIDs, then `_warn_uncovered_working_tree` warns, then the ack is written unconditionally. If a trigger file has staged content plus extra unstaged edits, the ack certifies staged content that the verify commands never ran on exactly.
-
-**Why deferred:** the warn-not-fail behavior is deliberate and documented (`docs/contribute/drift-contracts.md` — the digest is index-based by design, and `drift ack` warns on matching unstaged/untracked files). The failure mode is narrow: stage a trigger, edit it further without re-staging, ack, *and* have the staged-vs-tree delta flip a verify result. Codex's proposed blanket fail-on-any-unstaged-trigger contradicts the documented index-based philosophy and would add friction for the many contracts with no `verify_commands` at all.
-
-**Recorded candidate fix (if ever needed):** escalate warn→fail only when the contract has `verify_commands` AND a matching trigger file is unstaged/untracked — that is the one case where the verify guarantee is meaningfully weakened. Test approach: temp-repo test asserting `drift ack` raises `DriftAckError` for a verify-command contract with a dirty matching trigger, while still only warning for a no-verify contract.
diff --git a/wip/drift-hunt/deferred-code-findings.md b/wip/drift-hunt/deferred-code-findings.md
new file mode 100644
index 000000000..ba6dfc13d
--- /dev/null
+++ b/wip/drift-hunt/deferred-code-findings.md
@@ -0,0 +1,107 @@
+# Drift Hunt — deferred code findings
+
+Defects the hunt surfaced whose fix belongs in **code**, not docs (Louis' Checkpoint 0 ruling, 2026-07-12: doc-side these are fine; code-side issues get deferred here rather than fixed mid-campaign). One entry per finding; pick these up as ordinary follow-ups after the campaign (or sooner if one becomes pressing).
+
+> **Campaign closed 2026-07-12 — this file is the live handoff.** The per-section findings files and prescreen artifacts referenced below (`suspects.md`, the findings files) were pruned at close-out; they live in git history under the pushed `drift-hunt-archive` tag: `git show drift-hunt-archive:wip/drift-hunt/<path>`. The campaign record is `findings/SUMMARY.md` + `inventory.md`.
+
+## 1. `pipelex validate` shortcut forms are invisible in `--help`
+
+**Found:** Stage 0, via the CLI pre-screen near-miss (see `suspects.md` FP pattern 8).
+
+**Behavior:** `pipelex/cli/commands/validate/app.py` defines a custom click group whose `parse_args` rewrites `pipelex validate <pipe_code>` and `pipelex validate --all …` into `pipelex validate pipe …` — but the `--help`/`-h` path deliberately bypasses the rewrite, and the group's help output lists only the `method`/`pipe`/`bundle` subcommands with no mention of the shortcut. `docs/tools/cli/validate.md` documents the shortcut; the CLI's own help contradicts it.
+
+**Why it matters:** any consumer that derives the command surface from `--help` (a human skimming, an agent, tooling like the hunt's pre-screen) concludes the documented forms are dead. That's exactly the false "confirmed defect" Stage 0 produced before live execution disproved it.
+
+**Candidate fix (code):** surface the shortcut in the group help — e.g. an epilog/help line on the `validate` group ("`pipelex validate [PIPE_CODE|--all]` is a shortcut for `validate pipe …`"), or expose the forwarded options on the group so `--help` lists them. Keep the rewrite semantics unchanged; this is a help-surface fix only. Update `docs/tools/cli/validate.md` only if the chosen wording changes the documented contract.
+
+## 2. `plxt lint` does not detect invalid concept references (cross-repo: `vscode-pipelex`)
+
+**Found:** Stage 1 Part 3, as the code side of finding F10 (`docs/features/plxt.md:24`).
+
+**Behavior:** `plxt lint` (v0.6.0) is a pure JSON-Schema structural validation pass (required / type / enum / additionalProperties) plus a type-discriminated per-pipe-blueprint check. There is **no cross-reference resolution step**, so a `.mthds` file referencing a concept that was never declared lints **clean**: `output = "NonExistentConcept"` → exit 0, no diagnostics; `inputs = { topic = "TotallyMadeUpConcept" }` → exit 0. (Control: an unknown *pipe type* IS caught — `error[schema]: "UnknownPipeType" is not one of [...]`, exit 1.) The only concept-reference resolution in the toolchain is `crates/taplo-lsp/src/handlers/mthds_resolution.rs`, which is LSP-only (hover / goto-definition) and emits no diagnostic; `taplo-lsp/src/diagnostics.rs` has no concept logic at all.
+
+**Why it matters:** it is a capability gap, not a bug — but the docs promised it, which is what makes it worth recording. Undefined concept references are caught today only by `pipelex validate`, i.e. only once a human/agent reaches for the runtime. A fast, editor-time diagnostic is exactly what the linter is for.
+
+**Doc-side (in scope, Stage 2):** drop "or invalid concept references" from `docs/features/plxt.md:24` — the claim is currently false. That fix stands on its own and does **not** depend on the code side.
+
+**Candidate fix (code, other repo — `vscode-pipelex`):** add a concept cross-reference resolution pass to the lint path, reusing the resolution logic that already exists for the LSP, and surface it as a diagnostic in both `plxt lint` and the LSP. Out of scope for this campaign (D8: another repo's code); hand off to the toolchain owner. If the team decides `plxt lint` should stay purely structural, then the doc fix is the *whole* fix — and that is a legitimate outcome.
+
+## 3. `GatewayTelemetryManagerInjectedError`'s message names config fields that do not exist
+
+**Found:** Stage 1 Part 4, as a byproduct of the false-negative hand-check on `docs/setup/telemetry.md` (the page itself is clean — this defect is in an error string, not in the docs).
+
+**Behavior:** `pipelex/system/pipelex_service/exceptions.py:126-134` builds a remediation message telling the user to configure `host`, `project_api_key`, `langfuse_enabled`, `otlp_endpoint` and `otlp_headers`. **None of those field names exist** in the telemetry config models. The real names are `endpoint` and `api_key` (`PostHogConfig`, `telemetry_config.py:77-96`), `enabled` (`LangfuseConfig`, `:107-115`), and `endpoint` / `headers` on the `[[otlp]]` table.
+
+**Why it matters:** this is the error a user hits when telemetry is misconfigured — the one moment the message is load-bearing. Every key it hands them is wrong, so following it verbatim cannot work. It is the mirror image of doc drift (the docs are right; the *code's own* user-facing text is stale), and no docs fix can reach it.
+
+**Candidate fix (code):** rewrite the message against the current field names, and ideally derive them from the pydantic models rather than restating them, so the message cannot drift again.
+
+## 4. `ImageRenderable` docstring attributes the field-iterating implementation to the wrong class
+
+**Found:** Stage 2, batch S2-3, while fixing finding F33 of `docs/under-the-hood/stuffartefact-and-image-rendering.md` (the doc-side fix is applied; this is the same stale attribution living in code).
+
+**Behavior:** the docstring in `pipelex/tools/jinja2/image_renderable.py:29` still says "StuffContent (base): iterates model fields" — but `StuffContent` has no `render_with_images`; the field-iterating implementation lives on `StructuredContent` (`pipelex/core/stuffs/structured_content.py:52-71`), and a plain `StuffContent` subclass does not satisfy the protocol.
+
+**Why it matters:** the docstring is the first thing a contributor reads when implementing the protocol; it steers them to override the wrong base class — exactly the mistake the doc page used to teach before F33 was fixed.
+
+**Candidate fix (code):** correct the docstring line to name `StructuredContent`. One-line comment fix; deferred only because Stage 2 is docs-only (D17).
+
+## 5. Two docstrings still name `OrchestratorProtocol.run` after the `execute` rename
+
+**Found:** Stage 2, batch S2-3, by the blind post-fix review of the batch commit (the doc-side rename — finding F27 — is fully applied; these are the same stale name living in code comments).
+
+**Behavior:** `pipelex/runtime_bridge/delivery_mode.py:6` ("Delivery is passed as a parameter to ``OrchestratorProtocol.run``") and `pipelex/plugins/bundle_validator_registry.py:54` ("unlike ``OrchestratorProtocol.run``") both reference the protocol's blocking method by its old name; the method is `execute` (`pipelex/plugins/orchestrator_registry.py:38`).
+
+**Why it matters:** these docstrings describe the SPI seam an out-of-tree orchestrator plugin compiles against — the exact audience the F27 page-wide doc rename was fixed for.
+
+**Candidate fix (code):** rename both references to `OrchestratorProtocol.execute`. Two one-line comment fixes; deferred only because Stage 2 is docs-only (D17).
+
+## 6. CSV codec accepts `datetime` fields but its own error message, docstring, and tests omit the type
+
+**Found:** Stage 2, batch S2-1, as the code side of fix-time finding F52 on `docs/building-methods/pipes/csv-input-and-output.md` (the doc-side fix — adding `datetime` to the accepted flat types — is applied; it was verified to round-trip live).
+
+**Behavior:** the flatness classifier accepts `datetime` — `pipelex/tools/tabular/csv_codec.py:52` `_FLAT_SCALAR_TYPES = frozenset({str, int, float, bool, date, datetime})` — and a `datetime`-typed field round-trips end-to-end (ISO string in via pydantic lax coercion at `:364`, identical ISO string out via `model_dump(mode="json")` at `:402`). But the `CsvFlatnessError` message (`:104-108`) and the `flat_field_names` docstring (`:97`) both enumerate only text/integer/number/boolean/date, and `tests/unit/pipelex/tools/tabular/test_csv_codec.py` has zero `datetime` coverage (only `date`).
+
+**Why it matters:** the codec's own user-facing error text contradicts what the codec accepts, and the doc page's "contract is pinned by tests" claim is not yet true for `datetime` — the coercion table on that page deliberately omits a `datetime` row until tests pin it.
+
+**Candidate fix (code):** decide whether `datetime` is intended-supported. If yes (the classifier comment "Scalar python types a CSV cell can round-trip" suggests so): add it to the error message + docstring enumerations and pin the round trip with tests (then the doc's coercion table can gain its row). If no: remove `datetime` from `_FLAT_SCALAR_TYPES` — and revert the doc bullet.
+
+## 7. String-declared concepts drop their `refines` while basic-blueprint concepts keep `refines="native.Text"`
+
+**Found:** Stage 2, batch S2-1, by the adversarial verifier of fix-time finding F53 while pinning down accurate replacement wording for "native Text concepts" on `docs/building-methods/pipes/index.md`.
+
+**Behavior:** a concept declared as a plain string goes through `ConceptDeclarationType.STRING` → `_handle_basic_blueprint`, which generates a Text-based structure class — but the STRING branch discards the returned refine string and builds the `Concept` with `refines=None` (`pipelex/core/concepts/concept_factory.py:511-523`), whereas the BASIC_BLUEPRINT branch keeps `refines="native.Text"` (`:527-537`). Two spellings of the same declaration produce concepts with different `refines` metadata.
+
+**Why it matters:** anything that branches on `refines` (compatibility checks, tooling, future docs) sees an inconsistency between equivalent declarations. No doc claim currently hinges on it — recorded so the asymmetry is a decision, not an accident.
+
+**Candidate fix (code):** make the STRING branch keep the same `refines` the BASIC_BLUEPRINT branch records (or document why the asymmetry is intended).
+
+## 8. `tomli>=2.3.0` floor permits a tomli that rejects the TOML 1.1 syntax the docs and kit rely on
+
+**Found:** Stage 2, batch S2-2, by the adversarial verifier while refuting a fix-time candidate against `docs/building-methods/configure-ai-llm-to-optimize-methods.md` (the doc's multi-line inline tables turned out to be VALID under the shipped loader — this is the code-side residue of that refutation).
+
+**Behavior:** all deck/bundle TOML goes through `load_toml_from_path` → `tomli` (`pipelex/tools/misc/toml_utils.py:40`, `pipelex/core/interpreter/interpreter.py:43,46`). The installed tomli 2.4.1 accepts TOML 1.1 features (newlines and trailing commas in inline tables), and both the docs' snippets and real `.mthds` content use them. But the pin is `tomli>=2.3.0` (`pyproject.toml:55`), and tomli 2.3.0 empirically REJECTS the same content ("Invalid initial character for a key part") — 2.4.0 is the first version that accepts it.
+
+**Why it matters:** any environment resolving tomli to exactly 2.3.0 (permitted by the constraint) fails to parse documented deck syntax and multi-line inline tables in `.mthds` bundles generally — a version-dependent parse surface for the language itself.
+
+**Candidate fix (code):** bump the floor to `tomli>=2.4.0`. Also worth a decision: whether TOML 1.1 inline-table syntax is officially part of the MTHDS surface (the docs and examples already assume it).
+
+## 9. Re-running `pipelex build structures` over a migrated concept silently emits an opaque same-named class that can shadow the hand-written one
+
+**Found:** Stage 2, batch S2-2, by the adversarial verifier while refuting a missing-warning candidate on `docs/building-methods/concepts/python-classes.md` (the page's own migration procedure makes no false claim — the hazard lives outside it, so it was ruled out of drift scope and recorded here).
+
+**Behavior:** after the documented migration (hand-written class, concept's inline structure removed from the `.mthds`), regenerating with `pipelex build structures` over a scope that includes the migrated bundle emits `class <Name>(StructuredContent)` — an opaque `extra="allow"` placeholder docstringed "Imprecise: concept declares no structure" — with the SAME bare name as the hand-written class. Class registration is name-keyed and first-registered-wins (`pipelex/system/registries/class_registry_utils.py:169-171`), so whichever class is discovered first silently wins.
+
+**Why it matters:** a user who keeps generated output inside a discovered library dir can have the opaque placeholder silently shadow (or race) their hand-written class — no error, just wrong structure at runtime.
+
+**Candidate fix:** either skip emitting placeholder classes for structureless concepts whose name is already registered, or make registration collisions loud; docs-side, the Migration Guide could gain a warning to exclude migrated bundles from `build structures` scope (deferred with the rest — D17 keeps Stage 2 docs-only, and adding a new warning admonition is new content).
+
+## 10. The shipped `plxt.toml` template's "built-in default" comment contradicts one of the values it labels
+
+**Found:** Stage 2, batch S2-4, by a fix agent's whole-page scan while fixing `docs/tools/plxt.md` (the `align_entries` default finding), confirmed at fix time against the taplo source.
+
+**Behavior:** `pipelex/kit/configs/plxt.toml:34-35` (and its `.pipelex/plxt.toml` mirror) introduces the global `[formatting]` section with "Every option is shown at its built-in default", yet sets `array_auto_collapse = false` while the actual built-in default is `true` (`vscode-pipelex/crates/taplo/src/formatter/mod.rs:161`, `impl Default for Options`).
+
+**Why it matters:** the comment is the template's contract with the reader — the docs page (`docs/tools/plxt.md`) and this very campaign's verification both leaned on it as ground truth for defaults. One mislabeled value poisons that trust: either the value should flip to `true`, or the comment should stop claiming universality (and if `false` is a deliberate Pipelex-preferred override, it belongs in a differently-labeled line).
+
+**Candidate fix (code):** decide intent for `array_auto_collapse` in the shipped template: if the built-in default is wanted, set it to `true`; if the override is wanted, move it out of the "shown at built-in default" block or reword the comment. Same edit in both the kit template and the repo's own `.pipelex/plxt.toml`.
diff --git a/wip/drift-hunt/drift-hunt-explainer.html b/wip/drift-hunt/drift-hunt-explainer.html
new file mode 100644
index 000000000..9d1aec255
--- /dev/null
+++ b/wip/drift-hunt/drift-hunt-explainer.html
@@ -0,0 +1,298 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>Killing documentation drift — the Drift Hunt &amp; drift contracts</title>
+<style>
+ :root {
+ --ink: #1f2430;
+ --muted: #5b6472;
+ --line: #d9dee6;
+ --bg: #ffffff;
+ --panel: #f6f8fa;
+ --accent: #0b6bcb;
+ --accent-soft: #e8f1fb;
+ --good: #1a7f37;
+ --warn: #9a6700;
+ --bad: #cf222e;
+ }
+ * { box-sizing: border-box; }
+ html, body { margin: 0; padding: 0; }
+ body {
+ font-family: ui-sans-serif, system-ui, -apple-system, "Segoe UI", Roboto, sans-serif;
+ color: var(--ink);
+ background: var(--bg);
+ line-height: 1.55;
+ font-size: 16px;
+ }
+ main { max-width: 960px; margin: 0 auto; padding: 2.5rem 1.5rem 4rem; }
+ h1 { font-size: 1.9rem; line-height: 1.2; margin: 0 0 .3rem; }
+ .subtitle { color: var(--muted); margin: 0 0 2rem; font-size: 1.05rem; }
+ h2 { font-size: 1.45rem; margin: 3rem 0 .8rem; padding-top: .5rem; border-top: 1px solid var(--line); }
+ h3 { font-size: 1.1rem; margin: 1.6rem 0 .5rem; }
+ p { margin: .6rem 0; }
+ code {
+ font-family: ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
+ font-size: .85em;
+ background: var(--panel);
+ border: 1px solid var(--line);
+ border-radius: 4px;
+ padding: .08em .35em;
+ white-space: nowrap;
+ }
+ pre code { white-space: pre; display: block; padding: .8em 1em; overflow-x: auto; }
+ .tldr {
+ background: var(--accent-soft);
+ border: 1px solid #bcd6f2;
+ border-left: 5px solid var(--accent);
+ border-radius: 8px;
+ padding: 1.1rem 1.4rem;
+ margin: 1.5rem 0 2rem;
+ }
+ .tldr h2 { border: none; margin: 0 0 .5rem; padding: 0; font-size: 1.15rem; letter-spacing: .04em; }
+ .tldr ul { margin: .4rem 0 0; padding-left: 1.2rem; }
+ .tldr li { margin: .45rem 0; }
+ .callout {
+ background: var(--panel);
+ border: 1px solid var(--line);
+ border-left: 4px solid var(--muted);
+ border-radius: 6px;
+ padding: .8rem 1.1rem;
+ margin: 1rem 0;
+ }
+ .callout.rule { border-left-color: var(--accent); background: var(--accent-soft); }
+ .callout.honest { border-left-color: var(--warn); }
+ figure { margin: 1.4rem 0; }
+ .diagram {
+ border: 1px solid var(--line);
+ border-radius: 8px;
+ background: #fff;
+ padding: 1rem;
+ overflow-x: auto;
+ }
+ .diagram svg { display: block; margin: 0 auto; height: auto; max-width: 100%; }
+ .diagram.wide svg { max-width: none; }
+ figcaption { color: var(--muted); font-size: .88rem; margin-top: .5rem; text-align: center; }
+ table { border-collapse: collapse; width: 100%; margin: 1rem 0; font-size: .92rem; }
+ th, td { border: 1px solid var(--line); padding: .45rem .7rem; text-align: left; vertical-align: top; }
+ th { background: var(--panel); font-weight: 600; }
+ tbody tr:nth-child(even) { background: #fbfcfd; }
+ td.num, th.num { text-align: right; font-variant-numeric: tabular-nums; }
+ .sev-break { color: var(--bad); font-weight: 600; }
+ .sev-wrong { color: var(--warn); font-weight: 600; }
+ .ok { color: var(--good); font-weight: 600; }
+ details {
+ border: 1px solid var(--line);
+ border-radius: 8px;
+ margin: .8rem 0;
+ background: var(--panel);
+ }
+ details > summary {
+ cursor: pointer;
+ padding: .7rem 1rem;
+ font-weight: 600;
+ list-style: none;
+ }
+ details > summary::-webkit-details-marker { display: none; }
+ details > summary::before { content: "▸ "; color: var(--accent); }
+ details[open] > summary::before { content: "▾ "; }
+ details > .body { padding: 0 1.1rem .9rem; background: #fff; border-top: 1px solid var(--line); border-radius: 0 0 8px 8px; }
+ ul, ol { padding-left: 1.4rem; }
+ li { margin: .3rem 0; }
+ .pill {
+ display: inline-block;
+ font-size: .75rem;
+ font-weight: 700;
+ letter-spacing: .05em;
+ text-transform: uppercase;
+ border-radius: 999px;
+ padding: .12em .7em;
+ margin-right: .4em;
+ vertical-align: middle;
+ }
+ .pill.campaign { background: #eef7ee; color: var(--good); border: 1px solid #bfe3bf; }
+ .pill.system { background: var(--accent-soft); color: var(--accent); border: 1px solid #bcd6f2; }
+ footer { margin-top: 3.5rem; padding-top: 1rem; border-top: 1px solid var(--line); color: var(--muted); font-size: .85rem; }
+ .twocol { display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; }
+ @media (max-width: 720px) { .twocol { grid-template-columns: 1fr; } }
+</style>
+</head>
+<body>
+<main>
+
+<h1>Killing documentation drift</h1>
+<p class="subtitle">The Drift Hunt campaign &amp; the drift-contracts system — a recap for the team · July&nbsp;2026</p>
+
+<div class="tldr">
+<h2>TL;DR</h2>
+<ul>
+ <li><strong>We did two complementary things about code↔doc drift.</strong> <span class="pill campaign">campaign</span>A one-off <strong>Drift Hunt</strong> swept every hand-written docs page against the current code and fixed everything that had drifted — <strong>104 pages, 1,439 claims checked, ~140 confirmed defects fixed</strong>, landed on <code>dev</code> through five reviewed PRs. <span class="sev-break">39 of those were “break-a-user”</span>: a reader following the page verbatim hit an error — or worse, a <em>silent no-op</em>. <span class="pill system">system</span>A standing <strong>drift-contracts</strong> gate now prevents <em>new</em> drift: declared code→doc review obligations, mechanically enforced in <code>make check</code> and CI.</li>
+ <li><strong>What changes for you day-to-day:</strong> if you touch a <em>trigger</em> file (the config model, CLI code, the keyword-only guard), <code>make check</code> fails with an <em>open contract</em>. Run <code>make drift-plan</code>, actually review the named doc pages, fix what's stale, <code>git add</code> your changes, then <code>make drift-ack CONTRACT=… RATIONALE="…"</code> and commit. There is deliberately <strong>no bypass flag</strong> — but <em>“reviewed, no doc change needed”</em> is a perfectly legitimate rationale and takes one minute.</li>
+ <li><strong>Trust level of the docs today:</strong> every fix was re-verified against the code by an independent adversarial agent before being applied, every changed example re-validated against the real parser, and each fix batch survived a blind no-context review. Honest caveat: the cookbook pages were <strong>deliberately not swept</strong> (deferred).</li>
+</ul>
+</div>
+
+<h2>The problem, in one sentence</h2>
+<p>Documentation rots silently: the code renames a method, a default flips, a CLI grows a flag — and the page that explains it keeps teaching the old world, with no error anywhere. The worst finding of the campaign made the cost concrete: a guide told users to implement observer hooks under names the protocol no longer has, so a user following it to the letter got a <em>silent no-op</em> — their monitoring code simply never ran, no exception, nothing.</p>
+
+<p>Think of it as a stock-and-flow problem. The <strong>stock</strong> is the drift already accumulated in the pages; the <strong>flow</strong> is the new drift every future code change can introduce. The campaign drained the stock; the contracts throttle the flow.</p>
+
+<figure>
+ <div class="diagram wide"><svg id="dg1" width="100%" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" class="flowchart" style="max-width: 1670.81px; background-color: transparent;" viewBox="0 0 1670.8095703125 179" role="graphics-document document" aria-roledescription="flowchart-v2"><style>#dg1{font-family:ui-sans-serif,system-ui,sans-serif;font-size:14px;fill:#000000;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#dg1 .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#dg1 .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#dg1 .error-icon{fill:#552222;}#dg1 .error-text{fill:#552222;stroke:#552222;}#dg1 .edge-thickness-normal{stroke-width:1px;}#dg1 .edge-thickness-thick{stroke-width:3.5px;}#dg1 .edge-pattern-solid{stroke-dasharray:0;}#dg1 .edge-thickness-invisible{stroke-width:0;fill:none;}#dg1 .edge-pattern-dashed{stroke-dasharray:3;}#dg1 .edge-pattern-dotted{stroke-dasharray:2;}#dg1 .marker{fill:#666;stroke:#666;}#dg1 .marker.cross{stroke:#666;}#dg1 svg{font-family:ui-sans-serif,system-ui,sans-serif;font-size:14px;}#dg1 p{margin:0;}#dg1 .label{font-family:ui-sans-serif,system-ui,sans-serif;color:#000000;}#dg1 .cluster-label text{fill:#333;}#dg1 .cluster-label span{color:#333;}#dg1 .cluster-label span p{background-color:transparent;}#dg1 .label text,#dg1 span{fill:#000000;color:#000000;}#dg1 .node rect,#dg1 .node circle,#dg1 .node ellipse,#dg1 .node polygon,#dg1 .node path{fill:#eee;stroke:#999;stroke-width:1px;}#dg1 .rough-node .label text,#dg1 .node .label text,#dg1 .image-shape .label,#dg1 .icon-shape .label{text-anchor:middle;}#dg1 .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#dg1 .rough-node .label,#dg1 .node .label,#dg1 .image-shape .label,#dg1 .icon-shape .label{text-align:center;}#dg1 .node.clickable{cursor:pointer;}#dg1 .root .anchor path{fill:#666!important;stroke-width:0;stroke:#666;}#dg1 .arrowheadPath{fill:#333333;}#dg1 .edgePath .path{stroke:#666;stroke-width:1px;}#dg1 .flowchart-link{stroke:#666;fill:none;}#dg1 .edgeLabel{background-color:white;text-align:center;}#dg1 .edgeLabel p{background-color:white;}#dg1 .edgeLabel rect{opacity:0.5;background-color:white;fill:white;}#dg1 .labelBkg{background-color:rgba(255, 255, 255, 0.5);}#dg1 .cluster rect{fill:hsl(0, 0%, 98.9215686275%);stroke:#707070;stroke-width:1px;}#dg1 .cluster text{fill:#333;}#dg1 .cluster span{color:#333;}#dg1 div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:ui-sans-serif,system-ui,sans-serif;font-size:12px;background:hsl(-160, 0%, 93.3333333333%);border:1px solid #707070;border-radius:2px;pointer-events:none;z-index:100;}#dg1 .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#000000;}#dg1 rect.text{fill:none;stroke-width:0;}#dg1 .icon-shape,#dg1 .image-shape{background-color:white;text-align:center;}#dg1 .icon-shape p,#dg1 .image-shape p{background-color:white;padding:2px;}#dg1 .icon-shape .label rect,#dg1 .image-shape .label rect{opacity:0.5;background-color:white;fill:white;}#dg1 .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#dg1 .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#dg1 .node .neo-node{stroke:#999;}#dg1 [data-look="neo"].node rect,#dg1 [data-look="neo"].cluster rect,#dg1 [data-look="neo"].node polygon{stroke:url(#dg1-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg1 [data-look="neo"].node path{stroke:url(#dg1-gradient);stroke-width:1px;}#dg1 [data-look="neo"].node .outer-path{filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg1 [data-look="neo"].node .neo-line path{stroke:#999;filter:none;}#dg1 [data-look="neo"].node circle{stroke:url(#dg1-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg1 [data-look="neo"].node circle .state-start{fill:#000000;}#dg1 [data-look="neo"].icon-shape .icon{fill:url(#dg1-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg1 [data-look="neo"].icon-shape .icon-neo path{stroke:url(#dg1-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg1 :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;}</style><g><marker id="dg1_flowchart-v2-pointEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 0 L 10 5 L 0 10 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-pointStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="4.5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 5 L 10 10 L 10 0 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-pointEnd-margin" class="marker flowchart-v2" viewBox="0 0 11.5 14" refX="11.5" refY="7" markerUnits="userSpaceOnUse" markerWidth="10.5" markerHeight="14" orient="auto"><path d="M 0 0 L 11.5 7 L 0 14 z" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-pointStart-margin" class="marker flowchart-v2" viewBox="0 0 11.5 14" refX="1" refY="7" markerUnits="userSpaceOnUse" markerWidth="11.5" markerHeight="14" orient="auto"><polygon points="0,7 11.5,14 11.5,0" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-circleEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="11" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-circleStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-1" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-circleEnd-margin" class="marker flowchart-v2" viewBox="0 0 10 10" refY="5" refX="12.25" markerUnits="userSpaceOnUse" markerWidth="14" markerHeight="14" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-circleStart-margin" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-2" refY="5" markerUnits="userSpaceOnUse" markerWidth="14" markerHeight="14" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-crossEnd" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="12" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-crossStart" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="-1" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"/></marker><marker id="dg1_flowchart-v2-crossEnd-margin" class="marker cross flowchart-v2" viewBox="0 0 15 15" refX="17.7" refY="7.5" markerUnits="userSpaceOnUse" markerWidth="12" markerHeight="12" orient="auto"><path d="M 1,1 L 14,14 M 1,14 L 14,1" class="arrowMarkerPath" style="stroke-width: 2.5;"/></marker><marker id="dg1_flowchart-v2-crossStart-margin" class="marker cross flowchart-v2" viewBox="0 0 15 15" refX="-3.5" refY="7.5" markerUnits="userSpaceOnUse" markerWidth="12" markerHeight="12" orient="auto"><path d="M 1,1 L 14,14 M 1,14 L 14,1" class="arrowMarkerPath" style="stroke-width: 2.5; stroke-dasharray: 1, 0;"/></marker><g class="root"><g class="clusters"><g class="cluster" id="dg1-STOCK" data-look="classic"><rect style="" x="8" y="8" width="663.734375" height="163"/><g class="cluster-label" transform="translate(201.9375, 8)"><foreignObject width="275.859375" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5;"><span class="nodeLabel"><p>The stock — Drift Hunt campaign (one-off)</p></span></div></foreignObject></g></g></g><g class="edgePaths"><path d="M203.922,89.5L208.089,89.5C212.255,89.5,220.589,89.5,228.255,89.5C235.922,89.5,242.922,89.5,246.422,89.5L249.922,89.5" id="dg1-L_A_B_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_A_B_0" data-points="W3sieCI6MjAzLjkyMTg3NSwieSI6ODkuNX0seyJ4IjoyMjguOTIxODc1LCJ5Ijo4OS41fSx7IngiOjI1My45MjE4NzUsInkiOjg5LjV9XQ==" data-look="classic" marker-end="url(#dg1_flowchart-v2-pointEnd)"/><path d="M453.219,89.5L457.385,89.5C461.552,89.5,469.885,89.5,477.552,89.5C485.219,89.5,492.219,89.5,495.719,89.5L499.219,89.5" id="dg1-L_B_C_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_B_C_0" data-points="W3sieCI6NDUzLjIxODc1LCJ5Ijo4OS41fSx7IngiOjQ3OC4yMTg3NSwieSI6ODkuNX0seyJ4Ijo1MDMuMjE4NzUsInkiOjg5LjV9XQ==" data-look="classic" marker-end="url(#dg1_flowchart-v2-pointEnd)"/><path d="M646.734,89.5L650.901,89.5C655.068,89.5,663.401,89.5,671.734,89.5C680.068,89.5,688.401,89.5,696.151,89.57C703.901,89.641,711.068,89.781,714.652,89.851L718.235,89.922" id="dg1-L_C_G_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_C_G_0" data-points="W3sieCI6NjQ2LjczNDM3NSwieSI6ODkuNX0seyJ4Ijo2NzEuNzM0Mzc1LCJ5Ijo4OS41fSx7IngiOjY5Ni43MzQzNzUsInkiOjg5LjV9LHsieCI6NzIyLjIzNDM3NDk5OTk5OTcsInkiOjkwLjAwMDAwMDAwMDAwMDAxfV0=" data-look="classic" marker-end="url(#dg1_flowchart-v2-pointEnd)"/><path d="M872.857,89.5L877.023,89.5C881.19,89.5,889.523,89.5,897.19,89.5C904.857,89.5,911.857,89.5,915.357,89.5L918.857,89.5" id="dg1-L_G_FLOW_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_G_FLOW_0" data-points="W3sieCI6ODcyLjg1NjUwNjM0NzY1NjIsInkiOjg5LjV9LHsieCI6ODk3Ljg1NjUwNjM0NzY1NjIsInkiOjg5LjV9LHsieCI6OTIyLjg1NjUwNjM0NzY1NjIsInkiOjg5LjV9XQ==" data-look="classic" marker-end="url(#dg1_flowchart-v2-pointEnd)"/></g><g class="edgeLabels"><g class="edgeLabel"><g class="label" data-id="L_A_B_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_B_C_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_C_G_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_G_FLOW_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g></g><g class="nodes"><g class="root" transform="translate(914.8565063476562, 10.5)"><g class="clusters"><g class="cluster" id="dg1-FLOW" data-look="classic"><rect style="" x="8" y="8" width="739.953125" height="142"/><g class="cluster-label" transform="translate(234.65625, 8)"><foreignObject width="286.640625" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5;"><span class="nodeLabel"><p>The flow — drift contracts (standing system)</p></span></div></foreignObject></g></g></g><g class="edgePaths"><path d="M208.922,79L215.172,79C221.422,79,233.922,79,245.755,79C257.589,79,268.755,79,274.339,79L279.922,79" id="dg1-L_D_E_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_D_E_0" data-points="W3sieCI6MjA4LjkyMTg3NSwieSI6Nzl9LHsieCI6MjQ2LjQyMTg3NSwieSI6Nzl9LHsieCI6MjgzLjkyMTg3NSwieSI6Nzl9XQ==" data-look="classic" marker-end="url(#dg1_flowchart-v2-pointEnd)"/><path d="M450.984,79L457.234,79C463.484,79,475.984,79,487.818,79C499.651,79,510.818,79,516.401,79L521.984,79" id="dg1-L_E_F_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_E_F_0" data-points="W3sieCI6NDUwLjk4NDM3NSwieSI6Nzl9LHsieCI6NDg4LjQ4NDM3NSwieSI6Nzl9LHsieCI6NTI1Ljk4NDM3NSwieSI6Nzl9XQ==" data-look="classic" marker-end="url(#dg1_flowchart-v2-pointEnd)"/></g><g class="edgeLabels"><g class="edgeLabel"><g class="label" data-id="L_D_E_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_E_F_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g></g><g class="nodes"><g class="node default" id="dg1-flowchart-D-4" data-look="classic" transform="translate(127.2109375, 79)"><rect class="basic label-container" style="" x="-81.7109375" y="-36" width="163.421875" height="72"/><g class="label" style="" transform="translate(-51.7109375, -21)"><rect/><foreignObject width="103.421875" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>code change on<br />a trigger file</p></span></div></foreignObject></g></g><g class="node default" id="dg1-flowchart-E-5" data-look="classic" transform="translate(367.453125, 79)"><rect class="basic label-container" style="" x="-83.53125" y="-36" width="167.0625" height="72"/><g class="label" style="" transform="translate(-53.53125, -21)"><rect/><foreignObject width="107.0625" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>contract opens<br />make check fails</p></span></div></foreignObject></g></g><g class="node default" id="dg1-flowchart-F-7" data-look="classic" transform="translate(618.21875, 79)"><rect class="basic label-container" style="" x="-92.234375" y="-36" width="184.46875" height="72"/><g class="label" style="" transform="translate(-62.234375, -21)"><rect/><foreignObject width="124.46875" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>review + ack<br />recorded on the PR</p></span></div></foreignObject></g></g></g></g><g class="node default" id="dg1-flowchart-A-0" data-look="classic" transform="translate(118.4609375, 89.5)"><rect class="basic label-container" style="" x="-85.4609375" y="-36" width="170.921875" height="72"/><g class="label" style="" transform="translate(-55.4609375, -21)"><rect/><foreignObject width="110.921875" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>104 hand-written<br />doc pages</p></span></div></foreignObject></g></g><g class="node default" id="dg1-flowchart-B-1" data-look="classic" transform="translate(353.5703125, 89.5)"><rect class="basic label-container" style="" x="-99.6484375" y="-46.5" width="199.296875" height="93"/><g class="label" style="" transform="translate(-69.6484375, -31.5)"><rect/><foreignObject width="139.296875" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>Stage 1<br />semantic sweep<br />1,439 claims checked</p></span></div></foreignObject></g></g><g class="node default" id="dg1-flowchart-C-3" data-look="classic" transform="translate(574.9765625, 89.5)"><rect class="basic label-container" style="" x="-71.7578125" y="-46.5" width="143.515625" height="93"/><g class="label" style="" transform="translate(-41.7578125, -31.5)"><rect/><foreignObject width="83.515625" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>Stage 2<br />4 fix batches<br />4 PRs to dev</p></span></div></foreignObject></g></g><g class="node default" id="dg1-flowchart-G-9" data-look="classic" transform="translate(797.2954406738281, 89.5)"><g class="basic label-container outer-path"><path d="M-57.5703125 -18 C-19.399924062617103 -18, 18.770464374765794 -18, 57.5703125 -18 C57.5703125 -18, 57.5703125 -18, 57.5703125 -18 C57.8038298235217 -17.992511553948557, 58.037347147043405 -17.98502310789711, 58.72357645965283 -17.963017069506055 C59.02999866077668 -17.93345688829025, 59.33642086190053 -17.90389670707445, 59.87210141032111 -17.85222024881843 C60.20599994479562 -17.798238123186398, 60.53989847927014 -17.74425599755437, 61.0111678166247 -17.668064825839178 C61.23940248648252 -17.615971751814538, 61.467637156340345 -17.5638786777899, 62.13609501037113 -17.41130753470253 C62.449578986098956 -17.31826713621748, 62.763062961826776 -17.225226737732427, 63.24226042442517 -17.083003446192034 C63.67294554820416 -16.924507339431724, 64.10363067198315 -16.766011232671413, 64.32511858782874 -16.68450163222839 C64.62263133613882 -16.552801563953626, 64.92014408444892 -16.421101495678865, 65.38021980411605 -16.217439622243546 C65.69954183646144 -16.050849522504482, 66.01886386880682 -15.884259422765416, 66.40322843607088 -15.683736674221011 C66.73075113868687 -15.485190487158153, 67.05827384130286 -15.286644300095293, 67.38994072178987 -15.085585888053133 C67.68651580563218 -14.878707948093192, 67.98309088947451 -14.67183000813325, 68.33630204884189 -14.425445193623219 C68.64089526095252 -14.182540212530776, 68.94548847306315 -13.939635231438334, 69.23842361554019 -13.706027250644421 C69.49373372726093 -13.47416136231503, 69.74904383898166 -13.242295473985637, 70.09259841086275 -12.930288301759097 C70.34361755179532 -12.67109042969148, 70.59463669272787 -12.411892557623863, 70.89531644735568 -12.101416024703706 C71.08004437484588 -11.884424089756862, 71.26477230233608 -11.667432154810019, 71.64327918442453 -11.222816433457204 C71.84018348677668 -10.958982606656287, 72.03708778912885 -10.69514877985537, 72.3334130827452 -10.298099882199057 C72.48711258224945 -10.06197617686598, 72.64081208175372 -9.825852471532901, 72.96288223409623 -9.331066229589453 C73.18789923698404 -8.931525755519834, 73.41291623987185 -8.531985281450215, 73.529100014714 -8.325689224335035 C73.67680490487083 -8.018976859546026, 73.82450979502765 -7.712264494757019, 74.02973971428463 -7.286100176203093 C74.15597314352107 -6.974301165431768, 74.2822065727575 -6.6625021546604435, 74.46274409689568 -6.216570979583546 C74.5808093557485 -5.860977231063258, 74.69887461460132 -5.505383482542972, 74.8263338546599 -5.121496559358589 C74.88954049523306 -4.8804622449218, 74.95274713580622 -4.63942793048501, 75.11901491927283 -4.005376811213661 C75.19187132837703 -3.631274548671068, 75.26472773748124 -3.2571722861284758, 75.3395846014601 -2.872798110600833 C75.37292562115125 -2.614211745805249, 75.40626664084242 -2.3556253810096655, 75.48713653308556 -1.728414466338275 C75.516049892566 -1.278065553986686, 75.54496325204644 -0.8277166416350972, 75.56106439161238 -0.576928396289793 C75.56106439161238 -0.22944099486408914, 75.56106439161238 0.11804640656161469, 75.56106439161238 0.5769283962897886 C75.53371813022788 1.0028684908581313, 75.50637186884339 1.4288085854264743, 75.48713653308556 1.7284144663382628 C75.42889517305052 2.180122953712629, 75.3706538130155 2.6318314410869954, 75.3395846014601 2.8727981106008205 C75.25353710734056 3.314633806760336, 75.16748961322102 3.7564695029198516, 75.11901491927283 4.005376811213657 C75.02173139394537 4.376361056233847, 74.9244478686179 4.747345301254037, 74.8263338546599 5.121496559358577 C74.72574643918678 5.424449828335385, 74.62515902371366 5.7274030973121945, 74.46274409689569 6.216570979583534 C74.36388788139188 6.460747743132859, 74.26503166588806 6.704924506682184, 74.02973971428463 7.286100176203088 C73.92748513634622 7.498433995189662, 73.82523055840781 7.7107678141762355, 73.529100014714 8.325689224335031 C73.31407903547614 8.707480746207022, 73.09905805623826 9.089272268079013, 72.96288223409623 9.33106622958945 C72.8269290872402 9.539926771691912, 72.69097594038418 9.748787313794372, 72.33341308274521 10.298099882199049 C72.08905498363421 10.625517473526989, 71.84469688452323 10.952935064854927, 71.64327918442454 11.2228164334572 C71.39030250925796 11.519977230888468, 71.13732583409138 11.817138028319738, 70.89531644735568 12.101416024703695 C70.71151018848423 12.291211077762979, 70.52770392961277 12.481006130822262, 70.09259841086276 12.930288301759093 C69.84975779721168 13.15082972186646, 69.6069171835606 13.371371141973825, 69.2384236155402 13.706027250644409 C68.9369943424274 13.94640907458203, 68.63556506931461 14.186790898519654, 68.33630204884189 14.425445193623217 C68.11218531957326 14.581779322104166, 67.88806859030464 14.738113450585116, 67.38994072178988 15.085585888053126 C67.08786832616741 15.268703949453911, 66.78579593054492 15.451822010854695, 66.40322843607088 15.68373667422101 C66.03355911951546 15.876592919883471, 65.66388980296004 16.069449165545933, 65.38021980411605 16.217439622243543 C65.05446834188764 16.36164013149826, 64.72871687965925 16.505840640752975, 64.32511858782874 16.68450163222839 C63.9503353717984 16.822425320185623, 63.575552155768065 16.96034900814286, 63.24226042442518 17.08300344619203 C62.979818981208155 17.1608946889556, 62.71737753799112 17.238785931719168, 62.13609501037115 17.411307534702527 C61.856802798325006 17.475054159550048, 61.57751058627887 17.53880078439757, 61.011167816624706 17.668064825839174 C60.77661623690134 17.70598530627418, 60.542064657177974 17.74390578670919, 59.87210141032112 17.85222024881843 C59.508462783650955 17.887300029793867, 59.14482415698079 17.922379810769304, 58.72357645965283 17.963017069506055 C58.37466393514872 17.974206015049873, 58.0257514106446 17.98539496059369, 57.5703125 18 C57.5703125 18, 57.5703125 18, 57.5703125 18 C14.253306750844878 18, -29.063698998310244 18, -57.5703125 18 C-57.88590932536605 17.989879424082467, -58.20150615073211 17.97975884816493, -58.723576459652826 17.963017069506055 C-58.9718590449819 17.939065547102857, -59.22014163031097 17.915114024699655, -59.872101410321115 17.85222024881843 C-60.20529930750535 17.798351396789965, -60.53849720468959 17.744482544761503, -61.0111678166247 17.668064825839174 C-61.30554107030759 17.60087605165106, -61.59991432399048 17.533687277462942, -62.13609501037113 17.41130753470253 C-62.51239111235175 17.29962482685323, -62.88868721433237 17.187942119003925, -63.24226042442516 17.083003446192038 C-63.64343333194901 16.935368107821557, -64.04460623947286 16.787732769451075, -64.32511858782874 16.68450163222839 C-64.70586652832553 16.515955813456824, -65.08661446882232 16.347409994685254, -65.38021980411604 16.217439622243546 C-65.76522595887683 16.016582157913057, -66.15023211363761 15.815724693582569, -66.40322843607088 15.683736674221013 C-66.60813831228405 15.559519103076486, -66.81304818849725 15.435301531931959, -67.38994072178987 15.085585888053139 C-67.76418406164896 14.824529934900506, -68.13842740150805 14.563473981747872, -68.33630204884189 14.42544519362322 C-68.66988902143738 14.15941846010047, -69.00347599403287 13.89339172657772, -69.23842361554019 13.706027250644423 C-69.49491763017393 13.473086173013792, -69.75141164480769 13.240145095383163, -70.09259841086275 12.930288301759106 C-70.26612301748595 12.751109898785488, -70.43964762410917 12.57193149581187, -70.89531644735568 12.101416024703703 C-71.05472867384418 11.914161352657652, -71.2141409003327 11.7269066806116, -71.64327918442453 11.222816433457206 C-71.85357696250716 10.941036569348187, -72.0638747405898 10.659256705239168, -72.3334130827452 10.29809988219906 C-72.49815158725201 10.045017300794157, -72.66289009175884 9.791934719389257, -72.96288223409623 9.33106622958946 C-73.15100449992791 8.997036097160253, -73.33912676575959 8.663005964731045, -73.529100014714 8.325689224335036 C-73.6907422112602 7.9900357446179155, -73.8523844078064 7.6543822649007955, -74.02973971428463 7.286100176203095 C-74.17202500389506 6.934652759966394, -74.31431029350549 6.583205343729693, -74.46274409689568 6.2165709795835475 C-74.6041720492394 5.790612520608358, -74.74560000158313 5.364654061633167, -74.8263338546599 5.121496559358584 C-74.89556063227509 4.857504853236612, -74.96478740989029 4.593513147114639, -75.11901491927281 4.005376811213663 C-75.1810804842548 3.686683252592943, -75.2431460492368 3.3679896939722225, -75.3395846014601 2.8727981106008347 C-75.3701863229583 2.635457201654164, -75.40078804445652 2.3981162927074933, -75.48713653308556 1.7284144663382852 C-75.50822436970522 1.3999543891918635, -75.52931220632486 1.0714943120454419, -75.56106439161238 0.5769283962897952 C-75.56106439161238 0.1566133978758636, -75.56106439161238 -0.263701600538068, -75.56106439161238 -0.5769283962897864 C-75.53963609693483 -0.9106913825008128, -75.51820780225727 -1.2444543687118392, -75.48713653308556 -1.7284144663382603 C-75.43392691834615 -2.141097730830258, -75.38071730360674 -2.5537809953222554, -75.3395846014601 -2.8727981106008262 C-75.26617314623874 -3.2497504176459375, -75.1927616910174 -3.626702724691049, -75.11901491927283 -4.005376811213655 C-75.0035341527847 -4.4457550235137235, -74.88805338629655 -4.886133235813791, -74.8263338546599 -5.121496559358575 C-74.7034585904487 -5.491577277648018, -74.5805833262375 -5.861657995937461, -74.46274409689569 -6.2165709795835244 C-74.36039280332525 -6.4693806535301475, -74.25804150975482 -6.7221903274767705, -74.02973971428463 -7.286100176203086 C-73.87186136884543 -7.6139379385630175, -73.71398302340623 -7.941775700922949, -73.529100014714 -8.32568922433503 C-73.38543794494854 -8.580775776565648, -73.24177587518307 -8.835862328796269, -72.96288223409623 -9.331066229589455 C-72.81846886070235 -9.552923951542965, -72.6740554873085 -9.774781673496475, -72.33341308274521 -10.298099882199054 C-72.13091970067701 -10.569422566643834, -71.92842631860881 -10.840745251088615, -71.64327918442454 -11.2228164334572 C-71.35393610306305 -11.562695280912143, -71.06459302170155 -11.902574128367085, -70.89531644735568 -12.101416024703695 C-70.57879869174823 -12.428246593959404, -70.2622809361408 -12.75507716321511, -70.09259841086276 -12.930288301759086 C-69.89541608516035 -13.109364072803572, -69.69823375945793 -13.288439843848057, -69.2384236155402 -13.706027250644407 C-68.95932129937026 -13.928603920565376, -68.68021898320033 -14.151180590486344, -68.3363020488419 -14.425445193623215 C-68.01162113360085 -14.651928544885076, -67.68694021835981 -14.878411896146936, -67.38994072178988 -15.085585888053124 C-67.19165447030876 -15.205788178960336, -66.99336821882764 -15.325990469867548, -66.40322843607089 -15.683736674221008 C-66.08580098821837 -15.849338370443684, -65.76837354036584 -16.01494006666636, -65.38021980411605 -16.217439622243543 C-65.06907663750322 -16.355173472368943, -64.75793347089036 -16.49290732249434, -64.32511858782874 -16.684501632228386 C-63.90993694545095 -16.837292314520443, -63.49475530307316 -16.9900829968125, -63.24226042442518 -17.08300344619203 C-62.88830844696817 -17.18805453516943, -62.534356469511145 -17.29310562414683, -62.13609501037115 -17.411307534702523 C-61.766373451885364 -17.49569406776826, -61.39665189339958 -17.580080600834, -61.011167816624706 -17.668064825839174 C-60.77508929425391 -17.7062321705198, -60.53901077188312 -17.744399515200428, -59.87210141032112 -17.85222024881843 C-59.41894261620269 -17.895935932683848, -58.96578382208426 -17.939651616549266, -58.72357645965283 -17.963017069506055 C-58.452447962844964 -17.971711632679078, -58.181319466037095 -17.9804061958521, -57.57031250000001 -18 C-57.57031250000001 -18, -57.57031250000001 -18, -57.5703125 -18" stroke="none" stroke-width="0" fill="#eee" style=""/><path d="M-57.5703125 -18 C-15.289581321113815 -18, 26.99114985777237 -18, 57.5703125 -18 M-57.5703125 -18 C-19.36785552642418 -18, 18.83460144715164 -18, 57.5703125 -18 M57.5703125 -18 C57.5703125 -18, 57.5703125 -18, 57.5703125 -18 M57.5703125 -18 C57.5703125 -18, 57.5703125 -18, 57.5703125 -18 M57.5703125 -18 C57.86593873522029 -17.990519842038044, 58.16156497044058 -17.98103968407609, 58.72357645965283 -17.963017069506055 M57.5703125 -18 C57.952282181508565 -17.987750975772904, 58.33425186301714 -17.975501951545805, 58.72357645965283 -17.963017069506055 M58.72357645965283 -17.963017069506055 C59.16648013562751 -17.92029068455438, 59.60938381160219 -17.877564299602703, 59.87210141032111 -17.85222024881843 M58.72357645965283 -17.963017069506055 C59.05700528837583 -17.930851591405983, 59.39043411709883 -17.89868611330591, 59.87210141032111 -17.85222024881843 M59.87210141032111 -17.85222024881843 C60.115551869751954 -17.81286106652416, 60.3590023291828 -17.77350188422989, 61.0111678166247 -17.668064825839178 M59.87210141032111 -17.85222024881843 C60.324237617244194 -17.77912237325176, 60.77637382416727 -17.70602449768509, 61.0111678166247 -17.668064825839178 M61.0111678166247 -17.668064825839178 C61.338950347343534 -17.593250602183513, 61.66673287806237 -17.518436378527852, 62.13609501037113 -17.41130753470253 M61.0111678166247 -17.668064825839178 C61.42588275164798 -17.573408848187977, 61.84059768667127 -17.478752870536777, 62.13609501037113 -17.41130753470253 M62.13609501037113 -17.41130753470253 C62.41392479541623 -17.328849112859444, 62.691754580461335 -17.24639069101636, 63.24226042442517 -17.083003446192034 M62.13609501037113 -17.41130753470253 C62.434272497501745 -17.322810021687733, 62.73244998463236 -17.234312508672932, 63.24226042442517 -17.083003446192034 M63.24226042442517 -17.083003446192034 C63.57237675594936 -16.961517584629217, 63.902493087473545 -16.8400317230664, 64.32511858782874 -16.68450163222839 M63.24226042442517 -17.083003446192034 C63.610374743944746 -16.947533973797363, 63.978489063464316 -16.812064501402695, 64.32511858782874 -16.68450163222839 M64.32511858782874 -16.68450163222839 C64.61335710740853 -16.55690698990231, 64.90159562698832 -16.429312347576225, 65.38021980411605 -16.217439622243546 M64.32511858782874 -16.68450163222839 C64.58412384232857 -16.569847689179213, 64.84312909682838 -16.455193746130035, 65.38021980411605 -16.217439622243546 M65.38021980411605 -16.217439622243546 C65.64082724973163 -16.08148088479294, 65.90143469534719 -15.945522147342329, 66.40322843607088 -15.683736674221011 M65.38021980411605 -16.217439622243546 C65.59777387690532 -16.10394180181942, 65.81532794969458 -15.990443981395295, 66.40322843607088 -15.683736674221011 M66.40322843607088 -15.683736674221011 C66.79692129539073 -15.44507774930527, 67.19061415471057 -15.206418824389528, 67.38994072178987 -15.085585888053133 M66.40322843607088 -15.683736674221011 C66.68506070945867 -15.512888293863055, 66.96689298284646 -15.342039913505097, 67.38994072178987 -15.085585888053133 M67.38994072178987 -15.085585888053133 C67.66753527421426 -14.891947945156378, 67.94512982663863 -14.698310002259625, 68.33630204884189 -14.425445193623219 M67.38994072178987 -15.085585888053133 C67.61094723166484 -14.931421313981973, 67.8319537415398 -14.777256739910813, 68.33630204884189 -14.425445193623219 M68.33630204884189 -14.425445193623219 C68.58997413933857 -14.223148451949948, 68.84364622983524 -14.020851710276677, 69.23842361554019 -13.706027250644421 M68.33630204884189 -14.425445193623219 C68.59477460867627 -14.219320205398978, 68.85324716851066 -14.013195217174738, 69.23842361554019 -13.706027250644421 M69.23842361554019 -13.706027250644421 C69.450031955604 -13.51385015504833, 69.66164029566782 -13.321673059452237, 70.09259841086275 -12.930288301759097 M69.23842361554019 -13.706027250644421 C69.57269939300713 -13.402446834001356, 69.90697517047407 -13.09886641735829, 70.09259841086275 -12.930288301759097 M70.09259841086275 -12.930288301759097 C70.34997587517822 -12.66452493877058, 70.60735333949367 -12.398761575782064, 70.89531644735568 -12.101416024703706 M70.09259841086275 -12.930288301759097 C70.257942568858 -12.759556883621173, 70.42328672685325 -12.58882546548325, 70.89531644735568 -12.101416024703706 M70.89531644735568 -12.101416024703706 C71.18268522116786 -11.763856311237465, 71.47005399498003 -11.426296597771223, 71.64327918442453 -11.222816433457204 M70.89531644735568 -12.101416024703706 C71.11554526916288 -11.842722719234, 71.33577409097009 -11.584029413764295, 71.64327918442453 -11.222816433457204 M71.64327918442453 -11.222816433457204 C71.79792742572906 -11.015601878853758, 71.95257566703357 -10.808387324250312, 72.3334130827452 -10.298099882199057 M71.64327918442453 -11.222816433457204 C71.79925833407493 -11.013818582918518, 71.95523748372534 -10.80482073237983, 72.3334130827452 -10.298099882199057 M72.3334130827452 -10.298099882199057 C72.50555052429326 -10.03365054660602, 72.67768796584132 -9.769201211012986, 72.96288223409623 -9.331066229589453 M72.3334130827452 -10.298099882199057 C72.52338978119825 -10.006244657488732, 72.71336647965131 -9.714389432778406, 72.96288223409623 -9.331066229589453 M72.96288223409623 -9.331066229589453 C73.15818488515976 -8.984286596088259, 73.35348753622327 -8.637506962587064, 73.529100014714 -8.325689224335035 M72.96288223409623 -9.331066229589453 C73.16764453046409 -8.967490037957436, 73.37240682683196 -8.603913846325419, 73.529100014714 -8.325689224335035 M73.529100014714 -8.325689224335035 C73.64999239339815 -8.074653613314881, 73.7708847720823 -7.823618002294729, 74.02973971428463 -7.286100176203093 M73.529100014714 -8.325689224335035 C73.6816483318869 -8.008919379714397, 73.83419664905982 -7.692149535093762, 74.02973971428463 -7.286100176203093 M74.02973971428463 -7.286100176203093 C74.15116090238658 -6.986187494072299, 74.27258209048853 -6.686274811941505, 74.46274409689568 -6.216570979583546 M74.02973971428463 -7.286100176203093 C74.1436578198658 -7.004720252960881, 74.25757592544694 -6.7233403297186705, 74.46274409689568 -6.216570979583546 M74.46274409689568 -6.216570979583546 C74.59842453445458 -5.807923119398123, 74.73410497201348 -5.3992752592127005, 74.8263338546599 -5.121496559358589 M74.46274409689568 -6.216570979583546 C74.5793146059902 -5.8654791791791805, 74.69588511508475 -5.514387378774815, 74.8263338546599 -5.121496559358589 M74.8263338546599 -5.121496559358589 C74.93598916042208 -4.703333353577244, 75.04564446618426 -4.285170147795899, 75.11901491927283 -4.005376811213661 M74.8263338546599 -5.121496559358589 C74.91014588978386 -4.8018849450650976, 74.99395792490783 -4.482273330771607, 75.11901491927283 -4.005376811213661 M75.11901491927283 -4.005376811213661 C75.17827806882656 -3.701073063526375, 75.23754121838027 -3.396769315839089, 75.3395846014601 -2.872798110600833 M75.11901491927283 -4.005376811213661 C75.18599344282579 -3.6614562486851803, 75.25297196637875 -3.3175356861567, 75.3395846014601 -2.872798110600833 M75.3395846014601 -2.872798110600833 C75.38175806374709 -2.545709065195101, 75.42393152603407 -2.218620019789369, 75.48713653308556 -1.728414466338275 M75.3395846014601 -2.872798110600833 C75.39198066267873 -2.4664246065616795, 75.44437672389736 -2.060051102522526, 75.48713653308556 -1.728414466338275 M75.48713653308556 -1.728414466338275 C75.50194991127493 -1.4976841438951547, 75.51676328946431 -1.2669538214520342, 75.56106439161238 -0.576928396289793 M75.48713653308556 -1.728414466338275 C75.51525765285609 -1.2904053269529159, 75.54337877262662 -0.8523961875675569, 75.56106439161238 -0.576928396289793 M75.56106439161238 -0.576928396289793 C75.56106439161238 -0.2835786926064697, 75.56106439161238 0.00977101107685363, 75.56106439161238 0.5769283962897886 M75.56106439161238 -0.576928396289793 C75.56106439161238 -0.1356168032922579, 75.56106439161238 0.3056947897052772, 75.56106439161238 0.5769283962897886 M75.56106439161238 0.5769283962897886 C75.54420309901275 0.8395566387193181, 75.52734180641313 1.1021848811488475, 75.48713653308556 1.7284144663382628 M75.56106439161238 0.5769283962897886 C75.54611912138235 0.8097130437274831, 75.53117385115232 1.0424976911651775, 75.48713653308556 1.7284144663382628 M75.48713653308556 1.7284144663382628 C75.42818742261919 2.1856121262969577, 75.36923831215282 2.6428097862556523, 75.3395846014601 2.8727981106008205 M75.48713653308556 1.7284144663382628 C75.4306277519478 2.166685413921856, 75.37411897081002 2.604956361505449, 75.3395846014601 2.8727981106008205 M75.3395846014601 2.8727981106008205 C75.2584152655336 3.2895854954382124, 75.1772459296071 3.7063728802756044, 75.11901491927283 4.005376811213657 M75.3395846014601 2.8727981106008205 C75.2748752317474 3.205067047745451, 75.21016586203469 3.5373359848900807, 75.11901491927283 4.005376811213657 M75.11901491927283 4.005376811213657 C75.0365318005843 4.319920691182803, 74.95404868189576 4.634464571151949, 74.8263338546599 5.121496559358577 M75.11901491927283 4.005376811213657 C75.03685968822683 4.3186703134944455, 74.95470445718081 4.631963815775233, 74.8263338546599 5.121496559358577 M74.8263338546599 5.121496559358577 C74.7452925018374 5.365580201873867, 74.6642511490149 5.609663844389158, 74.46274409689569 6.216570979583534 M74.8263338546599 5.121496559358577 C74.70089927067318 5.499285541039652, 74.57546468668647 5.877074522720727, 74.46274409689569 6.216570979583534 M74.46274409689569 6.216570979583534 C74.31307699952137 6.5862516037145875, 74.16340990214705 6.95593222784564, 74.02973971428463 7.286100176203088 M74.46274409689569 6.216570979583534 C74.30361955268557 6.6096116802463065, 74.14449500847545 7.00265238090908, 74.02973971428463 7.286100176203088 M74.02973971428463 7.286100176203088 C73.92080909585385 7.512296936116447, 73.81187847742308 7.738493696029806, 73.529100014714 8.325689224335031 M74.02973971428463 7.286100176203088 C73.8641083257097 7.630037298522941, 73.69847693713474 7.973974420842794, 73.529100014714 8.325689224335031 M73.529100014714 8.325689224335031 C73.33790952098455 8.665167306088494, 73.1467190272551 9.004645387841958, 72.96288223409623 9.33106622958945 M73.529100014714 8.325689224335031 C73.31307475296975 8.709263951491732, 73.0970494912255 9.092838678648434, 72.96288223409623 9.33106622958945 M72.96288223409623 9.33106622958945 C72.76836637113935 9.629894829913972, 72.57385050818249 9.928723430238495, 72.33341308274521 10.298099882199049 M72.96288223409623 9.33106622958945 C72.75746325815383 9.646644939355525, 72.55204428221143 9.9622236491216, 72.33341308274521 10.298099882199049 M72.33341308274521 10.298099882199049 C72.19059492324004 10.489463204933983, 72.04777676373487 10.680826527668918, 71.64327918442454 11.2228164334572 M72.33341308274521 10.298099882199049 C72.1791174035166 10.504842035739843, 72.02482172428797 10.71158418928064, 71.64327918442454 11.2228164334572 M71.64327918442454 11.2228164334572 C71.35264468520064 11.564212253817086, 71.06201018597675 11.905608074176971, 70.89531644735568 12.101416024703695 M71.64327918442454 11.2228164334572 C71.3907929423192 11.519401140308307, 71.13830670021386 11.815985847159412, 70.89531644735568 12.101416024703695 M70.89531644735568 12.101416024703695 C70.65926529379185 12.345158220918492, 70.423214140228 12.588900417133289, 70.09259841086276 12.930288301759093 M70.89531644735568 12.101416024703695 C70.68268259909941 12.320977930591903, 70.47004875084315 12.54053983648011, 70.09259841086276 12.930288301759093 M70.09259841086276 12.930288301759093 C69.91902568969246 13.08792245464342, 69.74545296852216 13.245556607527748, 69.2384236155402 13.706027250644409 M70.09259841086276 12.930288301759093 C69.83054710034247 13.168276368475881, 69.5684957898222 13.40626443519267, 69.2384236155402 13.706027250644409 M69.2384236155402 13.706027250644409 C69.02378529629296 13.877195598478531, 68.80914697704571 14.048363946312653, 68.33630204884189 14.425445193623217 M69.2384236155402 13.706027250644409 C69.03327082636034 13.869631140670357, 68.82811803718046 14.033235030696305, 68.33630204884189 14.425445193623217 M68.33630204884189 14.425445193623217 C68.12355159872547 14.573850697805884, 67.91080114860904 14.722256201988552, 67.38994072178988 15.085585888053126 M68.33630204884189 14.425445193623217 C68.10460935893818 14.587063984285148, 67.87291666903448 14.748682774947078, 67.38994072178988 15.085585888053126 M67.38994072178988 15.085585888053126 C67.10169276745955 15.260323491901179, 66.81344481312921 15.43506109574923, 66.40322843607088 15.68373667422101 M67.38994072178988 15.085585888053126 C67.1630812697358 15.22310942106243, 66.93622181768171 15.360632954071736, 66.40322843607088 15.68373667422101 M66.40322843607088 15.68373667422101 C66.11541212559096 15.833890259043237, 65.82759581511104 15.984043843865464, 65.38021980411605 16.217439622243543 M66.40322843607088 15.68373667422101 C66.09574042363845 15.844152973351541, 65.788252411206 16.004569272482073, 65.38021980411605 16.217439622243543 M65.38021980411605 16.217439622243543 C65.07004601428339 16.3547443580239, 64.75987222445072 16.492049093804255, 64.32511858782874 16.68450163222839 M65.38021980411605 16.217439622243543 C65.10090208565069 16.341085290542434, 64.82158436718532 16.464730958841322, 64.32511858782874 16.68450163222839 M64.32511858782874 16.68450163222839 C63.909091820401656 16.837603328350838, 63.49306505297458 16.990705024473282, 63.24226042442518 17.08300344619203 M64.32511858782874 16.68450163222839 C63.97083224963199 16.814882279671803, 63.61654591143523 16.945262927115216, 63.24226042442518 17.08300344619203 M63.24226042442518 17.08300344619203 C62.88353279473535 17.18947192368422, 62.52480516504552 17.295940401176413, 62.13609501037115 17.411307534702527 M63.24226042442518 17.08300344619203 C62.92280577069976 17.177815910566043, 62.60335111697434 17.27262837494005, 62.13609501037115 17.411307534702527 M62.13609501037115 17.411307534702527 C61.77149304163571 17.494525554816192, 61.406891072900265 17.57774357492986, 61.011167816624706 17.668064825839174 M62.13609501037115 17.411307534702527 C61.80289089324062 17.4873592000775, 61.46968677611009 17.563410865452475, 61.011167816624706 17.668064825839174 M61.011167816624706 17.668064825839174 C60.59761617561949 17.734924648016023, 60.184064534614286 17.80178447019287, 59.87210141032112 17.85222024881843 M61.011167816624706 17.668064825839174 C60.74827132344034 17.710567892064468, 60.485374830255985 17.75307095828976, 59.87210141032112 17.85222024881843 M59.87210141032112 17.85222024881843 C59.52295098416489 17.8859023705149, 59.17380055800867 17.919584492211367, 58.72357645965283 17.963017069506055 M59.87210141032112 17.85222024881843 C59.47046155187383 17.890965962920237, 59.06882169342654 17.92971167702205, 58.72357645965283 17.963017069506055 M58.72357645965283 17.963017069506055 C58.26591293967194 17.97769344780384, 57.808249419691045 17.992369826101626, 57.5703125 18 M58.72357645965283 17.963017069506055 C58.35602284051545 17.97480379867207, 57.98846922137807 17.98659052783809, 57.5703125 18 M57.5703125 18 C57.5703125 18, 57.5703125 18, 57.5703125 18 M57.5703125 18 C57.5703125 18, 57.5703125 18, 57.5703125 18 M57.5703125 18 C16.802788120309998 18, -23.964736259380004 18, -57.5703125 18 M57.5703125 18 C23.524128747981074 18, -10.522055004037853 18, -57.5703125 18 M-57.5703125 18 C-57.88835705530096 17.98980093014767, -58.206401610601915 17.979601860295343, -58.723576459652826 17.963017069506055 M-57.5703125 18 C-57.813892768445704 17.992188854891175, -58.057473036891416 17.984377709782354, -58.723576459652826 17.963017069506055 M-58.723576459652826 17.963017069506055 C-59.18227749605661 17.91876673219289, -59.64097853246039 17.874516394879727, -59.872101410321115 17.85222024881843 M-58.723576459652826 17.963017069506055 C-59.128897403661426 17.9239162455073, -59.53421834767002 17.884815421508545, -59.872101410321115 17.85222024881843 M-59.872101410321115 17.85222024881843 C-60.10172974365151 17.815095720679892, -60.33135807698191 17.777971192541354, -61.0111678166247 17.668064825839174 M-59.872101410321115 17.85222024881843 C-60.11257378470714 17.813342540218812, -60.353046159093154 17.774464831619195, -61.0111678166247 17.668064825839174 M-61.0111678166247 17.668064825839174 C-61.280965211928695 17.60648533095358, -61.5507626072327 17.54490583606798, -62.13609501037113 17.41130753470253 M-61.0111678166247 17.668064825839174 C-61.38312304936905 17.583168471200008, -61.755078282113395 17.49827211656084, -62.13609501037113 17.41130753470253 M-62.13609501037113 17.41130753470253 C-62.38905505780104 17.336230320439718, -62.64201510523095 17.261153106176906, -63.24226042442516 17.083003446192038 M-62.13609501037113 17.41130753470253 C-62.47324780377587 17.31124235537113, -62.8104005971806 17.21117717603973, -63.24226042442516 17.083003446192038 M-63.24226042442516 17.083003446192038 C-63.53991427278184 16.97346407851477, -63.83756812113852 16.863924710837498, -64.32511858782874 16.68450163222839 M-63.24226042442516 17.083003446192038 C-63.46545882045294 17.00086437323046, -63.68865721648071 16.918725300268882, -64.32511858782874 16.68450163222839 M-64.32511858782874 16.68450163222839 C-64.6959728414545 16.52033545514883, -65.06682709508026 16.356169278069274, -65.38021980411604 16.217439622243546 M-64.32511858782874 16.68450163222839 C-64.7456711144055 16.498335503960313, -65.16622364098225 16.312169375692235, -65.38021980411604 16.217439622243546 M-65.38021980411604 16.217439622243546 C-65.59091893619623 16.107518019996544, -65.80161806827643 15.997596417749543, -66.40322843607088 15.683736674221013 M-65.38021980411604 16.217439622243546 C-65.70405574986955 16.048494616770764, -66.02789169562305 15.879549611297984, -66.40322843607088 15.683736674221013 M-66.40322843607088 15.683736674221013 C-66.65022878945318 15.534003608465468, -66.89722914283547 15.384270542709926, -67.38994072178987 15.085585888053139 M-66.40322843607088 15.683736674221013 C-66.65220909385181 15.53280313630146, -66.90118975163273 15.38186959838191, -67.38994072178987 15.085585888053139 M-67.38994072178987 15.085585888053139 C-67.5893657408691 14.9464756277134, -67.78879075994834 14.807365367373663, -68.33630204884189 14.42544519362322 M-67.38994072178987 15.085585888053139 C-67.74818197009223 14.8356923012511, -68.1064232183946 14.585798714449064, -68.33630204884189 14.42544519362322 M-68.33630204884189 14.42544519362322 C-68.64697177554483 14.17769435384626, -68.95764150224775 13.929943514069297, -69.23842361554019 13.706027250644423 M-68.33630204884189 14.42544519362322 C-68.52916324169483 14.271643524574884, -68.72202443454776 14.117841855526548, -69.23842361554019 13.706027250644423 M-69.23842361554019 13.706027250644423 C-69.54708932680529 13.425705218520642, -69.85575503807041 13.14538318639686, -70.09259841086275 12.930288301759106 M-69.23842361554019 13.706027250644423 C-69.44803803956262 13.51566097684099, -69.65765246358505 13.325294703037558, -70.09259841086275 12.930288301759106 M-70.09259841086275 12.930288301759106 C-70.3783052928487 12.635272489137371, -70.66401217483462 12.340256676515637, -70.89531644735568 12.101416024703703 M-70.09259841086275 12.930288301759106 C-70.33607312719279 12.678880667502368, -70.57954784352285 12.427473033245628, -70.89531644735568 12.101416024703703 M-70.89531644735568 12.101416024703703 C-71.11582201974498 11.842397632252718, -71.33632759213428 11.583379239801735, -71.64327918442453 11.222816433457206 M-70.89531644735568 12.101416024703703 C-71.0551161194593 11.913706236991864, -71.21491579156292 11.725996449280023, -71.64327918442453 11.222816433457206 M-71.64327918442453 11.222816433457206 C-71.89524105627267 10.885210479379438, -72.1472029281208 10.54760452530167, -72.3334130827452 10.29809988219906 M-71.64327918442453 11.222816433457206 C-71.90035058574654 10.878364175372159, -72.15742198706855 10.533911917287114, -72.3334130827452 10.29809988219906 M-72.3334130827452 10.29809988219906 C-72.46870902237362 10.09024898688306, -72.60400496200202 9.882398091567062, -72.96288223409623 9.33106622958946 M-72.3334130827452 10.29809988219906 C-72.54412909815971 9.974383497349562, -72.75484511357422 9.650667112500065, -72.96288223409623 9.33106622958946 M-72.96288223409623 9.33106622958946 C-73.17473858147898 8.954893832029969, -73.38659492886171 8.578721434470479, -73.529100014714 8.325689224335036 M-72.96288223409623 9.33106622958946 C-73.18024042846656 8.945124745581072, -73.39759862283691 8.559183261572684, -73.529100014714 8.325689224335036 M-73.529100014714 8.325689224335036 C-73.72916572719332 7.910248491651246, -73.92923143967265 7.494807758967456, -74.02973971428463 7.286100176203095 M-73.529100014714 8.325689224335036 C-73.69272012615764 7.985928562012771, -73.85634023760129 7.646167899690505, -74.02973971428463 7.286100176203095 M-74.02973971428463 7.286100176203095 C-74.16065928766548 6.962726298794613, -74.29157886104635 6.639352421386132, -74.46274409689568 6.2165709795835475 M-74.02973971428463 7.286100176203095 C-74.17980585350011 6.915433911009396, -74.32987199271561 6.544767645815698, -74.46274409689568 6.2165709795835475 M-74.46274409689568 6.2165709795835475 C-74.60603149514668 5.785012165798111, -74.74931889339769 5.353453352012676, -74.8263338546599 5.121496559358584 M-74.46274409689568 6.2165709795835475 C-74.57810829325324 5.8691124009431075, -74.6934724896108 5.521653822302667, -74.8263338546599 5.121496559358584 M-74.8263338546599 5.121496559358584 C-74.91612848482895 4.7790707172851015, -75.00592311499798 4.436644875211619, -75.11901491927281 4.005376811213663 M-74.8263338546599 5.121496559358584 C-74.92677427254051 4.738473714827288, -75.02721469042113 4.355450870295992, -75.11901491927281 4.005376811213663 M-75.11901491927281 4.005376811213663 C-75.19091161166763 3.636202491137633, -75.26280830406245 3.267028171061603, -75.3395846014601 2.8727981106008347 M-75.11901491927281 4.005376811213663 C-75.18928965310398 3.644530905560903, -75.25956438693514 3.283684999908143, -75.3395846014601 2.8727981106008347 M-75.3395846014601 2.8727981106008347 C-75.37144613749437 2.625686328890954, -75.40330767352864 2.3785745471810738, -75.48713653308556 1.7284144663382852 M-75.3395846014601 2.8727981106008347 C-75.37761529840296 2.5778395351226058, -75.41564599534581 2.282880959644377, -75.48713653308556 1.7284144663382852 M-75.48713653308556 1.7284144663382852 C-75.5133799754193 1.3196516686648685, -75.53962341775305 0.9108888709914519, -75.56106439161238 0.5769283962897952 M-75.48713653308556 1.7284144663382852 C-75.50382518264121 1.4684752786736959, -75.52051383219684 1.2085360910091067, -75.56106439161238 0.5769283962897952 M-75.56106439161238 0.5769283962897952 C-75.56106439161238 0.3110137703924376, -75.56106439161238 0.04509914449508001, -75.56106439161238 -0.5769283962897864 M-75.56106439161238 0.5769283962897952 C-75.56106439161238 0.12132356095126873, -75.56106439161238 -0.33428127438725774, -75.56106439161238 -0.5769283962897864 M-75.56106439161238 -0.5769283962897864 C-75.54557696690297 -0.8181575373629896, -75.53008954219356 -1.0593866784361927, -75.48713653308556 -1.7284144663382603 M-75.56106439161238 -0.5769283962897864 C-75.53855505707547 -0.9275294509099943, -75.51604572253855 -1.2781305055302021, -75.48713653308556 -1.7284144663382603 M-75.48713653308556 -1.7284144663382603 C-75.44823773035786 -2.0301058988569194, -75.40933892763016 -2.3317973313755784, -75.3395846014601 -2.8727981106008262 M-75.48713653308556 -1.7284144663382603 C-75.45171055549537 -2.003171352877251, -75.41628457790519 -2.2779282394162417, -75.3395846014601 -2.8727981106008262 M-75.3395846014601 -2.8727981106008262 C-75.28248389212041 -3.1659981848209524, -75.22538318278072 -3.4591982590410786, -75.11901491927283 -4.005376811213655 M-75.3395846014601 -2.8727981106008262 C-75.26426414782385 -3.2595527210394715, -75.1889436941876 -3.6463073314781163, -75.11901491927283 -4.005376811213655 M-75.11901491927283 -4.005376811213655 C-75.0071831591765 -4.431839780652344, -74.89535139908018 -4.858302750091032, -74.8263338546599 -5.121496559358575 M-75.11901491927283 -4.005376811213655 C-75.00619752066623 -4.435598447470467, -74.89338012205963 -4.865820083727278, -74.8263338546599 -5.121496559358575 M-74.8263338546599 -5.121496559358575 C-74.69147119008565 -5.527681417608664, -74.5566085255114 -5.933866275858754, -74.46274409689569 -6.2165709795835244 M-74.8263338546599 -5.121496559358575 C-74.69866020952966 -5.506029236457996, -74.57098656439942 -5.890561913557416, -74.46274409689569 -6.2165709795835244 M-74.46274409689569 -6.2165709795835244 C-74.36199979583715 -6.465411350956695, -74.2612554947786 -6.714251722329866, -74.02973971428463 -7.286100176203086 M-74.46274409689569 -6.2165709795835244 C-74.368826267595 -6.448549833733932, -74.27490843829432 -6.680528687884341, -74.02973971428463 -7.286100176203086 M-74.02973971428463 -7.286100176203086 C-73.897616922254 -7.5604559808295555, -73.76549413022337 -7.834811785456024, -73.529100014714 -8.32568922433503 M-74.02973971428463 -7.286100176203086 C-73.87815953311348 -7.600859665701291, -73.72657935194232 -7.915619155199497, -73.529100014714 -8.32568922433503 M-73.529100014714 -8.32568922433503 C-73.40434255896 -8.547208720088548, -73.27958510320599 -8.768728215842067, -72.96288223409623 -9.331066229589455 M-73.529100014714 -8.32568922433503 C-73.31024301565719 -8.714291987834677, -73.09138601660038 -9.102894751334325, -72.96288223409623 -9.331066229589455 M-72.96288223409623 -9.331066229589455 C-72.82716205934975 -9.539568863969885, -72.69144188460326 -9.748071498350313, -72.33341308274521 -10.298099882199054 M-72.96288223409623 -9.331066229589455 C-72.82220432180688 -9.54718528017466, -72.68152640951755 -9.763304330759864, -72.33341308274521 -10.298099882199054 M-72.33341308274521 -10.298099882199054 C-72.13255619553058 -10.567229812627787, -71.93169930831597 -10.83635974305652, -71.64327918442454 -11.2228164334572 M-72.33341308274521 -10.298099882199054 C-72.17576488997595 -10.509334098464048, -72.01811669720668 -10.720568314729045, -71.64327918442454 -11.2228164334572 M-71.64327918442454 -11.2228164334572 C-71.41826213106462 -11.487134268166807, -71.19324507770472 -11.751452102876414, -70.89531644735568 -12.101416024703695 M-71.64327918442454 -11.2228164334572 C-71.37731433052579 -11.535233884675572, -71.11134947662705 -11.847651335893945, -70.89531644735568 -12.101416024703695 M-70.89531644735568 -12.101416024703695 C-70.70027010356296 -12.302817388272324, -70.50522375977023 -12.50421875184095, -70.09259841086276 -12.930288301759086 M-70.89531644735568 -12.101416024703695 C-70.60633946299222 -12.39980848651061, -70.31736247862878 -12.698200948317524, -70.09259841086276 -12.930288301759086 M-70.09259841086276 -12.930288301759086 C-69.90460240072797 -13.10102130406165, -69.71660639059317 -13.271754306364214, -69.2384236155402 -13.706027250644407 M-70.09259841086276 -12.930288301759086 C-69.89421420429744 -13.110455589203234, -69.69582999773213 -13.290622876647383, -69.2384236155402 -13.706027250644407 M-69.2384236155402 -13.706027250644407 C-68.90616089239725 -13.970997930468508, -68.57389816925432 -14.235968610292609, -68.3363020488419 -14.425445193623215 M-69.2384236155402 -13.706027250644407 C-68.99490929334594 -13.900223442406071, -68.75139497115167 -14.094419634167734, -68.3363020488419 -14.425445193623215 M-68.3363020488419 -14.425445193623215 C-68.10037420087235 -14.590018247214838, -67.8644463529028 -14.75459130080646, -67.38994072178988 -15.085585888053124 M-68.3363020488419 -14.425445193623215 C-68.06112267978742 -14.61739840915273, -67.78594331073293 -14.809351624682245, -67.38994072178988 -15.085585888053124 M-67.38994072178988 -15.085585888053124 C-67.12494773334674 -15.246226194919013, -66.8599547449036 -15.4068665017849, -66.40322843607089 -15.683736674221008 M-67.38994072178988 -15.085585888053124 C-67.03504625658144 -15.300724998453859, -66.680151791373 -15.515864108854592, -66.40322843607089 -15.683736674221008 M-66.40322843607089 -15.683736674221008 C-66.08320318838057 -15.850693640955091, -65.76317794069026 -16.017650607689177, -65.38021980411605 -16.217439622243543 M-66.40322843607089 -15.683736674221008 C-66.09050929479038 -15.846882049920879, -65.77779015350987 -16.01002742562075, -65.38021980411605 -16.217439622243543 M-65.38021980411605 -16.217439622243543 C-65.02557963944813 -16.374428303115327, -64.6709394747802 -16.53141698398711, -64.32511858782874 -16.684501632228386 M-65.38021980411605 -16.217439622243543 C-65.10825803153732 -16.337829031504803, -64.83629625895857 -16.458218440766064, -64.32511858782874 -16.684501632228386 M-64.32511858782874 -16.684501632228386 C-63.99219439677685 -16.807020812052965, -63.65927020572497 -16.929539991877547, -63.24226042442518 -17.08300344619203 M-64.32511858782874 -16.684501632228386 C-63.98232916827575 -16.810651307334545, -63.639539748722754 -16.936800982440708, -63.24226042442518 -17.08300344619203 M-63.24226042442518 -17.08300344619203 C-62.99811516993777 -17.155464476179304, -62.753969915450355 -17.227925506166574, -62.13609501037115 -17.411307534702523 M-63.24226042442518 -17.08300344619203 C-62.861737326402526 -17.195940704278655, -62.48121422837986 -17.30887796236528, -62.13609501037115 -17.411307534702523 M-62.13609501037115 -17.411307534702523 C-61.694919690835704 -17.512002922448477, -61.25374437130026 -17.612698310194432, -61.011167816624706 -17.668064825839174 M-62.13609501037115 -17.411307534702523 C-61.72501792291781 -17.505133197385057, -61.313940835464464 -17.598958860067594, -61.011167816624706 -17.668064825839174 M-61.011167816624706 -17.668064825839174 C-60.597170748449386 -17.734996661226546, -60.18317368027406 -17.80192849661392, -59.87210141032112 -17.85222024881843 M-61.011167816624706 -17.668064825839174 C-60.740635836728494 -17.711802338341958, -60.47010385683228 -17.755539850844738, -59.87210141032112 -17.85222024881843 M-59.87210141032112 -17.85222024881843 C-59.626799001415286 -17.875884277186266, -59.38149659250944 -17.899548305554102, -58.72357645965283 -17.963017069506055 M-59.87210141032112 -17.85222024881843 C-59.474624199145474 -17.89056439734426, -59.077146987969826 -17.928908545870087, -58.72357645965283 -17.963017069506055 M-58.72357645965283 -17.963017069506055 C-58.422403383142274 -17.972675103871918, -58.121230306631716 -17.98233313823778, -57.57031250000001 -18 M-58.72357645965283 -17.963017069506055 C-58.41361385199414 -17.972956967028054, -58.10365124433545 -17.982896864550053, -57.57031250000001 -18 M-57.57031250000001 -18 C-57.57031250000001 -18, -57.57031250000001 -18, -57.5703125 -18 M-57.57031250000001 -18 C-57.57031250000001 -18, -57.5703125 -18, -57.5703125 -18" stroke="#999" stroke-width="1.3" fill="none" stroke-dasharray="0 0" style=""/></g><g class="label" style="" transform="translate(-63.5703125, -10.5)"><rect/><foreignObject width="127.140625" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>clean docs baseline</p></span></div></foreignObject></g></g></g></g></g><defs><filter id="dg1-drop-shadow" height="130%" width="130%"><feDropShadow dx="4" dy="4" stdDeviation="0" flood-opacity="0.06" flood-color="#000000"/></filter></defs><defs><filter id="dg1-drop-shadow-small" height="150%" width="150%"><feDropShadow dx="2" dy="2" stdDeviation="0" flood-opacity="0.06" flood-color="#000000"/></filter></defs><linearGradient id="dg1-gradient" gradientUnits="objectBoundingBox" x1="0%" y1="0%" x2="100%" y2="0%"><stop offset="0%" stop-color="hsl(0, 0%, 83.3333333333%)" stop-opacity="1"/><stop offset="100%" stop-color="hsl(0, 0%, 88.9215686275%)" stop-opacity="1"/></linearGradient></svg></div>
+ <figcaption>The campaign cleans up the past; the contracts guard the future.</figcaption>
+</figure>
+
+<h2><span class="pill campaign">campaign</span>Part 1 — the Drift Hunt</h2>
+
+<p>A campaign, not a feature: sweep every hand-written page under <code>docs/</code> (plus <code>CONTRIBUTING.md</code>, the agent instruction files, and friends) for statements the code can contradict — commands, flags, config keys, defaults, paths, API names, behaviors, and code snippets that should run. Editorial quality, tone, and structure were explicitly out of scope: a drift hunt, not a rewrite.</p>
+
+<figure>
+ <div class="diagram"><svg id="dg2" width="100%" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" class="flowchart" style="max-width: 374.484px; background-color: transparent;" viewBox="-14.2421875 0 374.484375 1210" role="graphics-document document" aria-roledescription="flowchart-v2"><style>#dg2{font-family:ui-sans-serif,system-ui,sans-serif;font-size:14px;fill:#000000;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#dg2 .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#dg2 .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#dg2 .error-icon{fill:#552222;}#dg2 .error-text{fill:#552222;stroke:#552222;}#dg2 .edge-thickness-normal{stroke-width:1px;}#dg2 .edge-thickness-thick{stroke-width:3.5px;}#dg2 .edge-pattern-solid{stroke-dasharray:0;}#dg2 .edge-thickness-invisible{stroke-width:0;fill:none;}#dg2 .edge-pattern-dashed{stroke-dasharray:3;}#dg2 .edge-pattern-dotted{stroke-dasharray:2;}#dg2 .marker{fill:#666;stroke:#666;}#dg2 .marker.cross{stroke:#666;}#dg2 svg{font-family:ui-sans-serif,system-ui,sans-serif;font-size:14px;}#dg2 p{margin:0;}#dg2 .label{font-family:ui-sans-serif,system-ui,sans-serif;color:#000000;}#dg2 .cluster-label text{fill:#333;}#dg2 .cluster-label span{color:#333;}#dg2 .cluster-label span p{background-color:transparent;}#dg2 .label text,#dg2 span{fill:#000000;color:#000000;}#dg2 .node rect,#dg2 .node circle,#dg2 .node ellipse,#dg2 .node polygon,#dg2 .node path{fill:#eee;stroke:#999;stroke-width:1px;}#dg2 .rough-node .label text,#dg2 .node .label text,#dg2 .image-shape .label,#dg2 .icon-shape .label{text-anchor:middle;}#dg2 .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#dg2 .rough-node .label,#dg2 .node .label,#dg2 .image-shape .label,#dg2 .icon-shape .label{text-align:center;}#dg2 .node.clickable{cursor:pointer;}#dg2 .root .anchor path{fill:#666!important;stroke-width:0;stroke:#666;}#dg2 .arrowheadPath{fill:#333333;}#dg2 .edgePath .path{stroke:#666;stroke-width:1px;}#dg2 .flowchart-link{stroke:#666;fill:none;}#dg2 .edgeLabel{background-color:white;text-align:center;}#dg2 .edgeLabel p{background-color:white;}#dg2 .edgeLabel rect{opacity:0.5;background-color:white;fill:white;}#dg2 .labelBkg{background-color:rgba(255, 255, 255, 0.5);}#dg2 .cluster rect{fill:hsl(0, 0%, 98.9215686275%);stroke:#707070;stroke-width:1px;}#dg2 .cluster text{fill:#333;}#dg2 .cluster span{color:#333;}#dg2 div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:ui-sans-serif,system-ui,sans-serif;font-size:12px;background:hsl(-160, 0%, 93.3333333333%);border:1px solid #707070;border-radius:2px;pointer-events:none;z-index:100;}#dg2 .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#000000;}#dg2 rect.text{fill:none;stroke-width:0;}#dg2 .icon-shape,#dg2 .image-shape{background-color:white;text-align:center;}#dg2 .icon-shape p,#dg2 .image-shape p{background-color:white;padding:2px;}#dg2 .icon-shape .label rect,#dg2 .image-shape .label rect{opacity:0.5;background-color:white;fill:white;}#dg2 .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#dg2 .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#dg2 .node .neo-node{stroke:#999;}#dg2 [data-look="neo"].node rect,#dg2 [data-look="neo"].cluster rect,#dg2 [data-look="neo"].node polygon{stroke:url(#dg2-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg2 [data-look="neo"].node path{stroke:url(#dg2-gradient);stroke-width:1px;}#dg2 [data-look="neo"].node .outer-path{filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg2 [data-look="neo"].node .neo-line path{stroke:#999;filter:none;}#dg2 [data-look="neo"].node circle{stroke:url(#dg2-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg2 [data-look="neo"].node circle .state-start{fill:#000000;}#dg2 [data-look="neo"].icon-shape .icon{fill:url(#dg2-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg2 [data-look="neo"].icon-shape .icon-neo path{stroke:url(#dg2-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg2 :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;}</style><g><marker id="dg2_flowchart-v2-pointEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 0 L 10 5 L 0 10 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-pointStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="4.5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 5 L 10 10 L 10 0 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-pointEnd-margin" class="marker flowchart-v2" viewBox="0 0 11.5 14" refX="11.5" refY="7" markerUnits="userSpaceOnUse" markerWidth="10.5" markerHeight="14" orient="auto"><path d="M 0 0 L 11.5 7 L 0 14 z" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-pointStart-margin" class="marker flowchart-v2" viewBox="0 0 11.5 14" refX="1" refY="7" markerUnits="userSpaceOnUse" markerWidth="11.5" markerHeight="14" orient="auto"><polygon points="0,7 11.5,14 11.5,0" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-circleEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="11" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-circleStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-1" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-circleEnd-margin" class="marker flowchart-v2" viewBox="0 0 10 10" refY="5" refX="12.25" markerUnits="userSpaceOnUse" markerWidth="14" markerHeight="14" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-circleStart-margin" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-2" refY="5" markerUnits="userSpaceOnUse" markerWidth="14" markerHeight="14" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-crossEnd" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="12" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-crossStart" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="-1" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"/></marker><marker id="dg2_flowchart-v2-crossEnd-margin" class="marker cross flowchart-v2" viewBox="0 0 15 15" refX="17.7" refY="7.5" markerUnits="userSpaceOnUse" markerWidth="12" markerHeight="12" orient="auto"><path d="M 1,1 L 14,14 M 1,14 L 14,1" class="arrowMarkerPath" style="stroke-width: 2.5;"/></marker><marker id="dg2_flowchart-v2-crossStart-margin" class="marker cross flowchart-v2" viewBox="0 0 15 15" refX="-3.5" refY="7.5" markerUnits="userSpaceOnUse" markerWidth="12" markerHeight="12" orient="auto"><path d="M 1,1 L 14,14 M 1,14 L 14,1" class="arrowMarkerPath" style="stroke-width: 2.5; stroke-dasharray: 1, 0;"/></marker><g class="root"><g class="clusters"><g class="cluster" id="dg2-S2" data-look="classic"><rect style="" x="8" y="773" width="330" height="307"/><g class="cluster-label" transform="translate(28.7578125, 773)"><foreignObject width="288.484375" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5;"><span class="nodeLabel"><p>Stage 2 — fix batches (4 parts, one PR each)</p></span></div></foreignObject></g></g><g class="cluster" id="dg2-S1" data-look="classic"><rect style="" x="-6.2421875" y="172" width="358.484375" height="429"/><g class="cluster-label" transform="translate(-2.2421875, 172)"><foreignObject width="350.484375" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5;"><span class="nodeLabel"><p>Stage 1 — semantic sweep (4 parts, one session each)</p></span></div></foreignObject></g></g></g><g class="edgePaths"><path d="M173,290L173,294.167C173,298.333,173,306.667,173,314.333C173,322,173,329,173,332.5L173,336" id="dg2-L_R_V_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_R_V_0" data-points="W3sieCI6MTczLCJ5IjoyOTB9LHsieCI6MTczLCJ5IjozMTV9LHsieCI6MTczLCJ5IjozNDB9XQ==" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/><path d="M173,433L173,437.167C173,441.333,173,449.667,173,457.333C173,465,173,472,173,475.5L173,479" id="dg2-L_V_FN_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_V_FN_0" data-points="W3sieCI6MTczLCJ5Ijo0MzN9LHsieCI6MTczLCJ5Ijo0NTh9LHsieCI6MTczLCJ5Ijo0ODN9XQ==" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/><path d="M173,576L173,580.167C173,584.333,173,592.667,173,601C173,609.333,173,617.667,173,625.333C173,633,173,640,173,643.5L173,647" id="dg2-L_FN_CP_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_FN_CP_0" data-points="W3sieCI6MTczLCJ5Ijo1NzZ9LHsieCI6MTczLCJ5Ijo2MDF9LHsieCI6MTczLCJ5Ijo2MjZ9LHsieCI6MTczLCJ5Ijo2NTF9XQ==" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/><path d="M173,912L173,916.167C173,920.333,173,928.667,173,936.333C173,944,173,951,173,954.5L173,958" id="dg2-L_FIX_BR_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_FIX_BR_0" data-points="W3sieCI6MTczLCJ5Ijo5MTJ9LHsieCI6MTczLCJ5Ijo5Mzd9LHsieCI6MTczLCJ5Ijo5NjJ9XQ==" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/><path d="M173,1055L173,1059.167C173,1063.333,173,1071.667,173,1080C173,1088.333,173,1096.667,173,1104.333C173,1112,173,1119,173,1122.5L173,1126" id="dg2-L_BR_PR_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_BR_PR_0" data-points="W3sieCI6MTczLCJ5IjoxMDU1fSx7IngiOjE3MywieSI6MTA4MH0seyJ4IjoxNzMsInkiOjExMDV9LHsieCI6MTczLCJ5IjoxMTMwfV0=" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/><path d="M173,122L173,126.167C173,130.333,173,138.667,173,146.333C173,154,173,161,173,164.5L173,168" id="dg2-L_S0_S1_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_S0_S1_0" data-points="W3sieCI6MTczLCJ5IjoxMjJ9LHsieCI6MTczLCJ5IjoxNDd9LHsieCI6MTczLCJ5IjoxNzJ9LHsieCI6MTczLCJ5IjoxOTd9XQ==" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/><path d="M173,723L173,727.167C173,731.333,173,739.667,173,747.333C173,755,173,762,173,765.5L173,769" id="dg2-L_CP_S2_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_CP_S2_0" data-points="W3sieCI6MTczLCJ5Ijo3MjN9LHsieCI6MTczLCJ5Ijo3NDh9LHsieCI6MTczLCJ5Ijo3NzN9LHsieCI6MTczLCJ5Ijo3OTh9XQ==" data-look="classic" marker-end="url(#dg2_flowchart-v2-pointEnd)"/></g><g class="edgeLabels"><g class="edgeLabel"><g class="label" data-id="L_R_V_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_V_FN_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_FN_CP_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_FIX_BR_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_BR_PR_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_S0_S1_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_CP_S2_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g></g><g class="nodes"><g class="node default" id="dg2-flowchart-S0-0" data-look="classic" transform="translate(173, 65)"><rect class="basic label-container" style="" x="-130" y="-57" width="260" height="114"/><g class="label" style="" transform="translate(-100, -42)"><rect/><foreignObject width="200" height="84"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table; white-space: break-spaces; line-height: 1.5; max-width: 200px; text-align: center; width: 200px;"><span class="nodeLabel"><p>Stage 0 — mechanical pre-screen<br />paths, config keys, CLI flags, TOML/Python parse</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-R-2" data-look="classic" transform="translate(173, 243.5)"><rect class="basic label-container" style="" x="-121.046875" y="-46.5" width="242.09375" height="93"/><g class="label" style="" transform="translate(-91.046875, -31.5)"><rect/><foreignObject width="182.09375" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>review agent per page:<br />extract checkable claims,<br />verify each against the code</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-V-3" data-look="classic" transform="translate(173, 386.5)"><rect class="basic label-container" style="" x="-106.9609375" y="-46.5" width="213.921875" height="93"/><g class="label" style="" transform="translate(-76.9609375, -31.5)"><rect/><foreignObject width="153.921875" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>adversarial verifier:<br />independent agent tries<br />to REFUTE each finding</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-FN-5" data-look="classic" transform="translate(173, 529.5)"><rect class="basic label-container" style="" x="-119.4140625" y="-46.5" width="238.828125" height="93"/><g class="label" style="" transform="translate(-89.4140625, -31.5)"><rect/><foreignObject width="178.828125" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>false-negative spot-check:<br />hand-verify CLEAN verdicts<br />on high-risk pages</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-CP-7" data-look="classic" transform="translate(173, 687)"><rect class="basic label-container" style="" x="-113.4375" y="-36" width="226.875" height="72"/><g class="label" style="" transform="translate(-83.4375, -21)"><rect/><foreignObject width="166.875" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>Checkpoint — Louis rules:<br />scope, severity, fix shape</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-FIX-10" data-look="classic" transform="translate(173, 855)"><rect class="basic label-container" style="" x="-130" y="-57" width="260" height="114"/><g class="label" style="" transform="translate(-100, -42)"><rect/><foreignObject width="200" height="84"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table; white-space: break-spaces; line-height: 1.5; max-width: 200px; text-align: center; width: 200px;"><span class="nodeLabel"><p>fix agents re-verify against code,<br />re-validate every example<br />against the real parser</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-BR-11" data-look="classic" transform="translate(173, 1008.5)"><rect class="basic label-container" style="" x="-104.296875" y="-46.5" width="208.59375" height="93"/><g class="label" style="" transform="translate(-74.296875, -31.5)"><rect/><foreignObject width="148.59375" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>blind post-fix review:<br />2 no-context reviewers<br />on the commit SHA</p></span></div></foreignObject></g></g><g class="node default" id="dg2-flowchart-PR-13" data-look="classic" transform="translate(173, 1166)"><rect class="basic label-container" style="" x="-83.046875" y="-36" width="166.09375" height="72"/><g class="label" style="" transform="translate(-53.046875, -21)"><rect/><foreignObject width="106.09375" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>4 docs-only PRs<br />merged to dev</p></span></div></foreignObject></g></g></g></g></g><defs><filter id="dg2-drop-shadow" height="130%" width="130%"><feDropShadow dx="4" dy="4" stdDeviation="0" flood-opacity="0.06" flood-color="#000000"/></filter></defs><defs><filter id="dg2-drop-shadow-small" height="150%" width="150%"><feDropShadow dx="2" dy="2" stdDeviation="0" flood-opacity="0.06" flood-color="#000000"/></filter></defs><linearGradient id="dg2-gradient" gradientUnits="objectBoundingBox" x1="0%" y1="0%" x2="100%" y2="0%"><stop offset="0%" stop-color="hsl(0, 0%, 83.3333333333%)" stop-opacity="1"/><stop offset="100%" stop-color="hsl(0, 0%, 88.9215686275%)" stop-opacity="1"/></linearGradient></svg></div>
+ <figcaption>The campaign pipeline. Every finding needed <code>file:line</code> evidence on both sides — the doc claim and the contradicting code — and had to survive an independent refutation attempt before anyone was allowed to fix it.</figcaption>
+</figure>
+
+<h3>The scoreboard</h3>
+
+<table>
+ <thead>
+ <tr><th>Section</th><th class="num">Pages</th><th class="num">Claims checked</th><th class="num">Confirmed drift</th><th class="num">Density (drift/page)</th></tr>
+ </thead>
+ <tbody>
+ <tr><td>building-methods (the language guide)</td><td class="num">35</td><td class="num">452</td><td class="num">44</td><td class="num">1.26</td></tr>
+ <tr><td>under-the-hood + advanced (internals)</td><td class="num">26</td><td class="num">435</td><td class="num">42</td><td class="num">1.62</td></tr>
+ <tr><td>features + reliability (capability pages)</td><td class="num">25</td><td class="num">308</td><td class="num">10</td><td class="num">0.40</td></tr>
+ <tr><td>the tail (get-started, contribute, tools, setup, root)</td><td class="num">18</td><td class="num">244</td><td class="num">13</td><td class="num">0.72</td></tr>
+ <tr><th>Total (Stage 1)</th><th class="num">104</th><th class="num">1,439</th><th class="num">109</th><th class="num">1.05</th></tr>
+ </tbody>
+</table>
+
+<p>Severity split of the 109: <span class="sev-break">39 break-a-user</span> (following the doc fails outright), <span class="sev-wrong">50 wrong</span>, 13 misleading, 7 stale details. Stage 2 then fixed <strong>all 109 with zero rejected at fix time</strong>, and — because fixers re-verify everything against the code — found and fixed <strong>31 more</strong> drift defects on the same pages, while <strong>refuting 13 candidates</strong> that didn't hold up. Four docs-only PRs carried the fixes to <code>dev</code>, each gated by strict docs build, full <code>make check</code>, and a blind post-fix review by two agents given nothing but the commit SHA.</p>
+
+<div class="callout honest"><strong>Honest coverage disclosure:</strong> Stage 1 covered 104 of the 125 in-scope pages. The 21 <code>docs/cookbook/</code> pages were <strong>never swept</strong> — that stage was deliberately deferred out of the campaign. No accuracy claim, positive or negative, applies to them.</div>
+
+<details>
+<summary>What actually drifts — the top defect classes</summary>
+<div class="body">
+<ol>
+ <li><strong>Renamed Python API never propagated into docs</strong> — the #1 class across all sections, and the source of the worst silent failures (mis-named observer hooks, renamed <code>execute</code>/<code>run</code> methods, renamed factory functions).</li>
+ <li><strong>Doc bundle examples that fail real validation</strong> — examples that parse as TOML but fail semantic validation (missing required fields, illegal inline sigils, operator constraints). Dominant in the language guide.</li>
+ <li><strong>Stale closed enumerations</strong> — a list written once (native concepts, the operator roster, model handles), the registry grows, the list is never revisited. One stale operator roster was copied on <em>three</em> pages, including the docs landing page.</li>
+ <li><strong>Wrong defaults / behavioral claims</strong> — flipped defaults, wrong fallback directories, wrong indexing bases, invented behaviors.</li>
+ <li><strong>Keyword-only convention drift in snippets</strong> — docs showing positional calls that now raise <code>TypeError</code> under the repo's keyword-only convention.</li>
+ <li><strong>Repo-tooling drift</strong> — the distinctive class of the contributor pages: not one phantom <code>make</code> target, but targets whose <em>behavior changed under a stable name</em> (e.g. a test target that grew heartbeat output weeks after a page was written on the premise it was silent).</li>
+ <li><strong>False prose beside a correct snippet</strong> — the most uncomfortable class: a tutorial whose examples all validate <em>and</em> dry-run, while the sentence explaining them teaches a false rule. A snippet checker returns green here; only an adversarial reader catches it.</li>
+</ol>
+</div>
+</details>
+
+<details>
+<summary>How we kept ourselves honest (the part worth stealing)</summary>
+<div class="body">
+<ul>
+ <li><strong>Adversarial verification before any action:</strong> no finding was acted on until a second, independent agent — prompted to <em>refute</em> it — confirmed it against the code with <code>file:line</code> evidence. Result: zero findings ever rejected downstream, so the pipeline's precision held.</li>
+ <li><strong>False-negative spot-checks:</strong> since nothing was ever refuted, the real risk was <em>under</em>-reporting. So on every part, the "clean" verdicts on high-risk pages were hand-verified — and once, that check broke a clean verdict both the reviewer and the verifier had passed (the false-prose-beside-correct-snippet case). It became a standing step.</li>
+ <li><strong>Blind post-fix reviews:</strong> after each fix batch, two reviewers with <em>no context</em> — just the commit SHA — re-reviewed the changes. They caught defects in three of the five reviewed batches, including one where the Stage-1 finding's own "correct" text was wrong. Rule learned: <em>the code is the authority, even over the finding that points at it.</em></li>
+ <li><strong>Run it, don't guess:</strong> live execution twice reversed a conclusion reached by reading — a CLI form "confirmed dead" from <code>--help</code> that a shortcut layer actually forwards, and a TOML construct "invalid" under a generic parser that the shipped loader accepts. Every changed example was re-validated against the real parser, load-bearing ones executed in dry-run.</li>
+ <li><strong>Validate the whole snippet, not just the fixed clause:</strong> a finding scoped to one kwarg masked a surrounding example that never ran at all. After that lesson, every touched snippet was import-checked or executed end to end.</li>
+</ul>
+</div>
+</details>
+
+<details>
+<summary>The headline lesson: drift follows <em>audience</em>, not complexity</summary>
+<div class="body">
+<p>The strongest pattern in the data is not <em>what</em> a page is about but <em>who it is for</em>. Pages whose author is also their reader — the maintainer-facing convention docs, the reliability pages written against the code — came back clean. Pages that explain something <em>to someone else</em> — onboarding, contributor setup, internals guides — drifted at up to 2 findings per page. And pages edited <strong>in the same commit as their code</strong> did not drift at all, which is precisely the behavior the contracts system forces. That correlation is where contract budget should go: reader-facing pages, not maintainer-facing ones.</p>
+</div>
+</details>
+
+<h2><span class="pill system">system</span>Part 2 — drift contracts, the standing gate</h2>
+
+<p>Shipped just before the campaign, then hardened. A <strong>drift contract</strong> declares: <em>when these trigger files change, these doc pages must be re-examined against the change, and the examination must be recorded.</em> The tool can't judge whether prose is still true — that takes a human or an agent — but it can prove, deterministically, that someone looked at <em>this exact content</em> and said so on the record.</p>
+
+<h3>Three tiers — contracts are only the last resort</h3>
+
+<table>
+ <thead><tr><th>Tier</th><th>Mechanism</th><th>Judgment needed</th></tr></thead>
+ <tbody>
+ <tr><td><strong>Derived</strong></td><td>Regenerate and diff (config sync, schema, gateway models, error pages)</td><td>None — output is a pure function of sources</td></tr>
+ <tr><td><strong>Linkage</strong></td><td>Referential integrity of declared cross-references (spec ↔ conformance links)</td><td>None — links resolve or they don't</td></tr>
+ <tr><td><strong>Review</strong></td><td><strong>Drift contracts</strong>: "this changed → that must be looked at, and the look recorded"</td><td>Yes — someone must judge whether the prose still holds</td></tr>
+ </tbody>
+</table>
+
+<p>Standing rule: anything mechanizable gets demoted into a derived check. The best review contract is one you eventually mechanize out of existence.</p>
+
+<h3>The one rule everything hangs on</h3>
+
+<div class="callout rule">
+<p><strong>A contract is <em>fulfilled</em> iff the stored ack digest equals the digest recomputed from the current tree. Otherwise it is <em>open</em>.</strong></p>
+<p>No base ref, no git diff, no timestamps — pure content equality. The digest covers the <code>(path, blob&nbsp;OID)</code> of every tracked file matching the contract's triggers, <em>plus the contract's own definition</em>. Edit a trigger file, add or delete a matched file, or edit the contract itself — it opens. And the digest reads the <strong>git index</strong>, not the working tree, which is what lets the ack land in the same commit as the change it certifies.</p>
+</div>
+
+<h3>The lifecycle, end to end</h3>
+
+<figure>
+ <div class="diagram"><svg id="dg3" width="100%" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" class="flowchart" style="max-width: 540.064px; background-color: transparent;" viewBox="0 0 540.0635375976562 1642.625" role="graphics-document document" aria-roledescription="flowchart-v2"><style>#dg3{font-family:ui-sans-serif,system-ui,sans-serif;font-size:14px;fill:#000000;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#dg3 .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#dg3 .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#dg3 .error-icon{fill:#552222;}#dg3 .error-text{fill:#552222;stroke:#552222;}#dg3 .edge-thickness-normal{stroke-width:1px;}#dg3 .edge-thickness-thick{stroke-width:3.5px;}#dg3 .edge-pattern-solid{stroke-dasharray:0;}#dg3 .edge-thickness-invisible{stroke-width:0;fill:none;}#dg3 .edge-pattern-dashed{stroke-dasharray:3;}#dg3 .edge-pattern-dotted{stroke-dasharray:2;}#dg3 .marker{fill:#666;stroke:#666;}#dg3 .marker.cross{stroke:#666;}#dg3 svg{font-family:ui-sans-serif,system-ui,sans-serif;font-size:14px;}#dg3 p{margin:0;}#dg3 .label{font-family:ui-sans-serif,system-ui,sans-serif;color:#000000;}#dg3 .cluster-label text{fill:#333;}#dg3 .cluster-label span{color:#333;}#dg3 .cluster-label span p{background-color:transparent;}#dg3 .label text,#dg3 span{fill:#000000;color:#000000;}#dg3 .node rect,#dg3 .node circle,#dg3 .node ellipse,#dg3 .node polygon,#dg3 .node path{fill:#eee;stroke:#999;stroke-width:1px;}#dg3 .rough-node .label text,#dg3 .node .label text,#dg3 .image-shape .label,#dg3 .icon-shape .label{text-anchor:middle;}#dg3 .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#dg3 .rough-node .label,#dg3 .node .label,#dg3 .image-shape .label,#dg3 .icon-shape .label{text-align:center;}#dg3 .node.clickable{cursor:pointer;}#dg3 .root .anchor path{fill:#666!important;stroke-width:0;stroke:#666;}#dg3 .arrowheadPath{fill:#333333;}#dg3 .edgePath .path{stroke:#666;stroke-width:1px;}#dg3 .flowchart-link{stroke:#666;fill:none;}#dg3 .edgeLabel{background-color:white;text-align:center;}#dg3 .edgeLabel p{background-color:white;}#dg3 .edgeLabel rect{opacity:0.5;background-color:white;fill:white;}#dg3 .labelBkg{background-color:rgba(255, 255, 255, 0.5);}#dg3 .cluster rect{fill:hsl(0, 0%, 98.9215686275%);stroke:#707070;stroke-width:1px;}#dg3 .cluster text{fill:#333;}#dg3 .cluster span{color:#333;}#dg3 div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:ui-sans-serif,system-ui,sans-serif;font-size:12px;background:hsl(-160, 0%, 93.3333333333%);border:1px solid #707070;border-radius:2px;pointer-events:none;z-index:100;}#dg3 .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#000000;}#dg3 rect.text{fill:none;stroke-width:0;}#dg3 .icon-shape,#dg3 .image-shape{background-color:white;text-align:center;}#dg3 .icon-shape p,#dg3 .image-shape p{background-color:white;padding:2px;}#dg3 .icon-shape .label rect,#dg3 .image-shape .label rect{opacity:0.5;background-color:white;fill:white;}#dg3 .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#dg3 .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#dg3 .node .neo-node{stroke:#999;}#dg3 [data-look="neo"].node rect,#dg3 [data-look="neo"].cluster rect,#dg3 [data-look="neo"].node polygon{stroke:url(#dg3-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg3 [data-look="neo"].node path{stroke:url(#dg3-gradient);stroke-width:1px;}#dg3 [data-look="neo"].node .outer-path{filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg3 [data-look="neo"].node .neo-line path{stroke:#999;filter:none;}#dg3 [data-look="neo"].node circle{stroke:url(#dg3-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg3 [data-look="neo"].node circle .state-start{fill:#000000;}#dg3 [data-look="neo"].icon-shape .icon{fill:url(#dg3-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg3 [data-look="neo"].icon-shape .icon-neo path{stroke:url(#dg3-gradient);filter:drop-shadow( 1px 2px 2px rgba(185,185,185,1));}#dg3 :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;}</style><g><marker id="dg3_flowchart-v2-pointEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 0 L 10 5 L 0 10 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-pointStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="4.5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 5 L 10 10 L 10 0 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-pointEnd-margin" class="marker flowchart-v2" viewBox="0 0 11.5 14" refX="11.5" refY="7" markerUnits="userSpaceOnUse" markerWidth="10.5" markerHeight="14" orient="auto"><path d="M 0 0 L 11.5 7 L 0 14 z" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-pointStart-margin" class="marker flowchart-v2" viewBox="0 0 11.5 14" refX="1" refY="7" markerUnits="userSpaceOnUse" markerWidth="11.5" markerHeight="14" orient="auto"><polygon points="0,7 11.5,14 11.5,0" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-circleEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="11" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-circleStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-1" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-circleEnd-margin" class="marker flowchart-v2" viewBox="0 0 10 10" refY="5" refX="12.25" markerUnits="userSpaceOnUse" markerWidth="14" markerHeight="14" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-circleStart-margin" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-2" refY="5" markerUnits="userSpaceOnUse" markerWidth="14" markerHeight="14" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 0; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-crossEnd" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="12" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-crossStart" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="-1" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"/></marker><marker id="dg3_flowchart-v2-crossEnd-margin" class="marker cross flowchart-v2" viewBox="0 0 15 15" refX="17.7" refY="7.5" markerUnits="userSpaceOnUse" markerWidth="12" markerHeight="12" orient="auto"><path d="M 1,1 L 14,14 M 1,14 L 14,1" class="arrowMarkerPath" style="stroke-width: 2.5;"/></marker><marker id="dg3_flowchart-v2-crossStart-margin" class="marker cross flowchart-v2" viewBox="0 0 15 15" refX="-3.5" refY="7.5" markerUnits="userSpaceOnUse" markerWidth="12" markerHeight="12" orient="auto"><path d="M 1,1 L 14,14 M 1,14 L 14,1" class="arrowMarkerPath" style="stroke-width: 2.5; stroke-dasharray: 1, 0;"/></marker><g class="root"><g class="clusters"/><g class="edgePaths"><path d="M229.857,80L229.857,84.167C229.857,88.333,229.857,96.667,229.857,104.333C229.857,112,229.857,119,229.857,122.5L229.857,126" id="dg3-L_E_C_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_E_C_0" data-points="W3sieCI6MjI5Ljg1NzIyMzUxMDc0MjIsInkiOjgwfSx7IngiOjIyOS44NTcyMjM1MTA3NDIyLCJ5IjoxMDV9LHsieCI6MjI5Ljg1NzIyMzUxMDc0MjIsInkiOjEzMH1d" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M175.702,306.954L162.522,321.896C149.342,336.839,122.983,366.724,109.881,391.75C96.78,416.776,96.936,436.943,97.014,447.026L97.093,457.109" id="dg3-L_C_OK_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_C_OK_0" data-points="W3sieCI6MTc1LjcwMTczNjg0MjUxMTc2LCJ5IjozMDYuOTUzODg4MzMxNzY5Nn0seyJ4Ijo5Ni42MjM1NjU2NzM4MjgxMiwieSI6Mzk2LjYwOTM3NX0seyJ4Ijo5Ny4xMjM1NjU2NzM4MjgxMiwieSI6NDYxLjEwOTM3NX1d" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M284.013,306.954L297.192,321.896C310.372,336.839,336.731,366.724,349.911,386.917C363.091,407.109,363.091,417.609,363.091,422.859L363.091,428.109" id="dg3-L_C_P_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_C_P_0" data-points="W3sieCI6Mjg0LjAxMjcxMDE3ODk3MjYsInkiOjMwNi45NTM4ODgzMzE3Njk2M30seyJ4IjozNjMuMDkwODgxMzQ3NjU2MjUsInkiOjM5Ni42MDkzNzV9LHsieCI6MzYzLjA5MDg4MTM0NzY1NjI1LCJ5Ijo0MzIuMTA5Mzc1fV0=" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M290.691,525.109L284.204,529.276C277.716,533.443,264.742,541.776,258.254,549.443C251.767,557.109,251.767,564.109,251.767,567.609L251.767,571.109" id="dg3-L_P_REV_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_P_REV_0" data-points="W3sieCI6MjkwLjY5MTIxNDYwODE0NTc0LCJ5Ijo1MjUuMTA5Mzc1fSx7IngiOjI1MS43NjY2NjI1OTc2NTYyNSwieSI6NTUwLjEwOTM3NX0seyJ4IjoyNTEuNzY2NjYyNTk3NjU2MjUsInkiOjU3NS4xMDkzNzV9XQ==" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M251.767,668.109L251.767,672.276C251.767,676.443,251.767,684.776,251.767,692.443C251.767,700.109,251.767,707.109,251.767,710.609L251.767,714.109" id="dg3-L_REV_ADD_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_REV_ADD_0" data-points="W3sieCI6MjUxLjc2NjY2MjU5NzY1NjI1LCJ5Ijo2NjguMTA5Mzc1fSx7IngiOjI1MS43NjY2NjI1OTc2NTYyNSwieSI6NjkzLjEwOTM3NX0seyJ4IjoyNTEuNzY2NjYyNTk3NjU2MjUsInkiOjcxOC4xMDkzNzV9XQ==" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M251.767,811.109L251.767,815.276C251.767,819.443,251.767,827.776,251.767,835.443C251.767,843.109,251.767,850.109,251.767,853.609L251.767,857.109" id="dg3-L_ADD_ACK_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_ADD_ACK_0" data-points="W3sieCI6MjUxLjc2NjY2MjU5NzY1NjI1LCJ5Ijo4MTEuMTA5Mzc1fSx7IngiOjI1MS43NjY2NjI1OTc2NTYyNSwieSI6ODM2LjEwOTM3NX0seyJ4IjoyNTEuNzY2NjYyNTk3NjU2MjUsInkiOjg2MS4xMDkzNzV9XQ==" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M251.767,996.109L251.767,1000.276C251.767,1004.443,251.767,1012.776,251.767,1020.443C251.767,1028.109,251.767,1035.109,251.767,1038.609L251.767,1042.109" id="dg3-L_ACK_COMMIT_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_ACK_COMMIT_0" data-points="W3sieCI6MjUxLjc2NjY2MjU5NzY1NjI1LCJ5Ijo5OTYuMTA5Mzc1fSx7IngiOjI1MS43NjY2NjI1OTc2NTYyNSwieSI6MTAyMS4xMDkzNzV9LHsieCI6MjUxLjc2NjY2MjU5NzY1NjI1LCJ5IjoxMDQ2LjEwOTM3NX1d" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M251.767,1118.109L251.767,1122.276C251.767,1126.443,251.767,1134.776,251.767,1142.443C251.767,1150.109,251.767,1157.109,251.767,1160.609L251.767,1164.109" id="dg3-L_COMMIT_PRV_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_COMMIT_PRV_0" data-points="W3sieCI6MjUxLjc2NjY2MjU5NzY1NjI1LCJ5IjoxMTE4LjEwOTM3NX0seyJ4IjoyNTEuNzY2NjYyNTk3NjU2MjUsInkiOjExNDMuMTA5Mzc1fSx7IngiOjI1MS43NjY2NjI1OTc2NTYyNSwieSI6MTE2OC4xMDkzNzV9XQ==" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M251.767,1261.109L251.767,1265.276C251.767,1269.443,251.767,1277.776,261.681,1293.81C271.595,1309.844,291.423,1333.579,301.337,1345.447L311.252,1357.314" id="dg3-L_PRV_MERGE_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_PRV_MERGE_0" data-points="W3sieCI6MjUxLjc2NjY2MjU5NzY1NjI1LCJ5IjoxMjYxLjEwOTM3NX0seyJ4IjoyNTEuNzY2NjYyNTk3NjU2MjUsInkiOjEyODYuMTA5Mzc1fSx7IngiOjMxMy44MTYxMzk3MDU1NDUyNSwieSI6MTM2MC4zODQxMTY2NDIxMTEyfV0=" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M363.091,1527.625L363.091,1533.542C363.091,1539.458,363.091,1551.292,363.165,1562.542C363.239,1573.792,363.387,1584.459,363.461,1589.792L363.535,1595.125" id="dg3-L_MERGE_DONE_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_MERGE_DONE_0" data-points="W3sieCI6MzYzLjA5MDg4MTM0NzY1NjI1LCJ5IjoxNTI3LjYyNX0seyJ4IjozNjMuMDkwODgxMzQ3NjU2MjUsInkiOjE1NjMuMTI1fSx7IngiOjM2My41OTA4ODEzNDc2NTYyNSwieSI6MTU5OS4xMjV9XQ==" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/><path d="M412.366,1360.384L422.707,1348.005C433.049,1335.626,453.732,1310.868,464.074,1286.572C474.415,1262.276,474.415,1238.443,474.415,1214.609C474.415,1190.776,474.415,1166.943,474.415,1144.859C474.415,1122.776,474.415,1102.443,474.415,1082.109C474.415,1061.776,474.415,1041.443,474.415,1015.859C474.415,990.276,474.415,959.443,474.415,928.609C474.415,897.776,474.415,866.943,474.415,839.609C474.415,812.276,474.415,788.443,474.415,764.609C474.415,740.776,474.415,716.943,474.415,693.109C474.415,669.276,474.415,645.443,474.415,621.609C474.415,597.776,474.415,573.943,468.489,558.22C462.562,542.497,450.709,534.884,444.783,531.077L438.856,527.271" id="dg3-L_MERGE_P_0" class="edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style=";" data-edge="true" data-et="edge" data-id="L_MERGE_P_0" data-points="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" data-look="classic" marker-end="url(#dg3_flowchart-v2-pointEnd)"/></g><g class="edgeLabels"><g class="edgeLabel"><g class="label" data-id="L_E_C_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel" transform="translate(96.62356567382812, 396.609375)"><g class="label" data-id="L_C_OK_0" transform="translate(-62.78125, -10.5)"><foreignObject width="125.5625" height="21"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"><p>digest matches ack</p></span></div></foreignObject></g></g><g class="edgeLabel" transform="translate(363.09088134765625, 396.609375)"><g class="label" data-id="L_C_P_0" transform="translate(-42.5859375, -10.5)"><foreignObject width="85.171875" height="21"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"><p>digest differs</p></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_P_REV_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_REV_ADD_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_ADD_ACK_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_ACK_COMMIT_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_COMMIT_PRV_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" data-id="L_PRV_MERGE_0" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"></span></div></foreignObject></g></g><g class="edgeLabel" transform="translate(363.09088134765625, 1563.125)"><g class="label" data-id="L_MERGE_DONE_0" transform="translate(-74.671875, -10.5)"><foreignObject width="149.34375" height="21"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"><p>covers merged content</p></span></div></foreignObject></g></g><g class="edgeLabel" transform="translate(474.41510009765625, 928.609375)"><g class="label" data-id="L_MERGE_P_0" transform="translate(-57.6484375, -10.5)"><foreignObject width="115.296875" height="21"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel"><p>spliced / stale ack</p></span></div></foreignObject></g></g></g><g class="nodes"><g class="node default" id="dg3-flowchart-E-0" data-look="classic" transform="translate(229.8572235107422, 44)"><rect class="basic label-container" style="" x="-103.3671875" y="-36" width="206.734375" height="72"/><g class="label" style="" transform="translate(-73.3671875, -21)"><rect/><foreignObject width="146.734375" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>you edit a trigger file<br />(e.g. the config model)</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-C-1" data-look="classic" transform="translate(229.8572235107422, 245.5546875)"><polygon points="115.5546875,0 231.109375,-115.5546875 115.5546875,-231.109375 0,-115.5546875" class="label-container" transform="translate(-115.0546875, 115.5546875)"/><g class="label" style="" transform="translate(-79.5546875, -21)"><rect/><foreignObject width="159.109375" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>make drift-check<br />(inside make check + CI)</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-OK-3" data-look="classic" transform="translate(96.62356567382812, 478.609375)"><g class="basic label-container outer-path"><path d="M-70.6328125 -18 C-21.092581844606315 -18, 28.44764881078737 -18, 70.6328125 -18 C70.6328125 -18, 70.6328125 -18, 70.6328125 -18 C71.08165243491854 -17.985606576901098, 71.53049236983708 -17.971213153802193, 71.78607645965283 -17.963017069506055 C72.17415140355988 -17.925579946395747, 72.5622263474669 -17.88814282328544, 72.93460141032111 -17.85222024881843 C73.35672608158877 -17.783974405593593, 73.77885075285643 -17.715728562368753, 74.07366781662469 -17.668064825839178 C74.47416187068521 -17.57665467146781, 74.87465592474572 -17.48524451709644, 75.19859501037112 -17.41130753470253 C75.41993462239414 -17.34561509968943, 75.64127423441717 -17.279922664676327, 76.30476042442517 -17.083003446192034 C76.57508070326692 -16.983523084765007, 76.84540098210867 -16.884042723337977, 77.38761858782874 -16.68450163222839 C77.6607773548102 -16.563582349052623, 77.93393612179165 -16.442663065876857, 78.44271980411605 -16.217439622243546 C78.79366273018611 -16.034352922502467, 79.14460565625616 -15.85126622276139, 79.46572843607088 -15.683736674221011 C79.80992186665488 -15.475084593895872, 80.15411529723889 -15.266432513570734, 80.45244072178987 -15.085585888053133 C80.68034498352357 -14.92660974088419, 80.90824924525725 -14.767633593715248, 81.39880204884189 -14.425445193623219 C81.69798730464329 -14.186852913775628, 81.99717256044468 -13.948260633928037, 82.30092361554019 -13.706027250644421 C82.53841336788383 -13.490345341200435, 82.77590312022748 -13.274663431756448, 83.15509841086275 -12.930288301759097 C83.32152157252945 -12.758442723742547, 83.48794473419616 -12.586597145725998, 83.95781644735568 -12.101416024703706 C84.1762793720069 -11.844797042515818, 84.39474229665811 -11.58817806032793, 84.70577918442453 -11.222816433457204 C84.91336911166593 -10.944664842785162, 85.12095903890732 -10.666513252113118, 85.3959130827452 -10.298099882199057 C85.56710292713352 -10.035106310308855, 85.73829277152183 -9.77211273841865, 86.02538223409623 -9.331066229589453 C86.23452082513417 -8.959719483622823, 86.44365941617212 -8.588372737656194, 86.591600014714 -8.325689224335035 C86.70565138919449 -8.088859104917827, 86.81970276367497 -7.85202898550062, 87.09223971428463 -7.286100176203093 C87.20625582066128 -7.004478189101305, 87.32027192703792 -6.722856201999516, 87.52524409689568 -6.216570979583546 C87.61031472336734 -5.960351806814553, 87.695385349839 -5.7041326340455605, 87.8888338546599 -5.121496559358589 C88.00148930700773 -4.691892494374361, 88.11414475935558 -4.262288429390133, 88.18151491927283 -4.005376811213661 C88.2479977856285 -3.6640013436360417, 88.31448065198418 -3.322625876058422, 88.4020846014601 -2.872798110600833 C88.43247462711405 -2.6370990728181374, 88.462864652768 -2.401400035035442, 88.54963653308556 -1.728414466338275 C88.57766092928699 -1.2919118746155718, 88.60568532548842 -0.8554092828928685, 88.62356439161238 -0.576928396289793 C88.62356439161238 -0.31623953830075135, 88.62356439161238 -0.05555068031170973, 88.62356439161238 0.5769283962897886 C88.59585310267992 1.0085540895625917, 88.56814181374746 1.4401797828353946, 88.54963653308556 1.7284144663382628 C88.5198896419135 1.9591254800807458, 88.49014275074143 2.1898364938232286, 88.4020846014601 2.8727981106008205 C88.32186725020496 3.2846972553527887, 88.24164989894982 3.696596400104757, 88.18151491927283 4.005376811213657 C88.0784692530373 4.3983345968439655, 87.97542358680178 4.791292382474274, 87.8888338546599 5.121496559358577 C87.79202032233843 5.41308349356922, 87.69520679001697 5.704670427779862, 87.52524409689569 6.216570979583534 C87.4282186483672 6.456225713308325, 87.33119319983872 6.695880447033116, 87.09223971428463 7.286100176203088 C86.96419500069139 7.551987763697414, 86.83615028709816 7.81787535119174, 86.591600014714 8.325689224335031 C86.46727726034308 8.546436863764598, 86.34295450597216 8.767184503194162, 86.02538223409623 9.33106622958945 C85.84307282321767 9.611142441630363, 85.6607634123391 9.89121865367128, 85.39591308274521 10.298099882199049 C85.24402616738743 10.501614511595823, 85.09213925202965 10.705129140992598, 84.70577918442454 11.2228164334572 C84.45337336493336 11.51930667132955, 84.20096754544215 11.815796909201898, 83.95781644735568 12.101416024703695 C83.72800859457475 12.338711500275615, 83.49820074179381 12.576006975847534, 83.15509841086276 12.930288301759093 C82.92498609750332 13.139270216313959, 82.69487378414388 13.348252130868822, 82.3009236155402 13.706027250644409 C82.04517881016481 13.909976927276194, 81.78943400478941 14.113926603907977, 81.39880204884189 14.425445193623217 C81.19950138448401 14.564468709492566, 81.00020072012613 14.703492225361915, 80.45244072178988 15.085585888053126 C80.1086514135941 15.293992987011634, 79.76486210539834 15.502400085970143, 79.46572843607088 15.68373667422101 C79.09621741561038 15.876510336912146, 78.72670639514989 16.069283999603282, 78.44271980411605 16.217439622243543 C78.04288300009318 16.39443551464709, 77.6430461960703 16.571431407050635, 77.38761858782874 16.68450163222839 C77.10650748637624 16.78795311612544, 76.82539638492376 16.891404600022494, 76.30476042442518 17.08300344619203 C75.9882828557094 17.17693232736222, 75.6718052869936 17.27086120853241, 75.19859501037115 17.411307534702527 C74.96875646959863 17.463766681797207, 74.73891792882613 17.516225828891887, 74.0736678166247 17.668064825839174 C73.64932749723488 17.736668877869796, 73.22498717784505 17.805272929900422, 72.93460141032112 17.85222024881843 C72.57097898822154 17.88729846655843, 72.20735656612196 17.922376684298435, 71.78607645965283 17.963017069506055 C71.37599919245399 17.976167449298142, 70.96592192525515 17.98931782909023, 70.6328125 18 C70.6328125 18, 70.6328125 18, 70.6328125 18 C24.636754508546943 18, -21.359303482906114 18, -70.6328125 18 C-70.95261615384435 17.989744519281263, -71.27241980768869 17.979489038562523, -71.78607645965282 17.963017069506055 C-72.10030349068272 17.932703965735897, -72.41453052171263 17.902390861965735, -72.93460141032111 17.85222024881843 C-73.27379766109877 17.797381629498908, -73.61299391187643 17.74254301017939, -74.0736678166247 17.668064825839174 C-74.42530454235438 17.587806037835456, -74.77694126808406 17.507547249831738, -75.19859501037112 17.41130753470253 C-75.42469697238427 17.344201659210448, -75.65079893439741 17.277095783718362, -76.30476042442517 17.083003446192038 C-76.67817363508745 16.945583932874438, -77.05158684574972 16.80816441955684, -77.38761858782874 16.68450163222839 C-77.64026728407042 16.572661548959488, -77.89291598031211 16.460821465690586, -78.44271980411604 16.217439622243546 C-78.83344990514387 16.013595978811903, -79.2241800061717 15.809752335380262, -79.46572843607088 15.683736674221013 C-79.82278486129738 15.467286970952244, -80.17984128652388 15.250837267683476, -80.45244072178987 15.085585888053139 C-80.73052016120887 14.891609708979358, -81.00859960062787 14.697633529905579, -81.39880204884189 14.42544519362322 C-81.66759838854841 14.211087265678271, -81.93639472825494 13.996729337733324, -82.30092361554019 13.706027250644423 C-82.6176096549567 13.418421368699793, -82.93429569437322 13.13081548675516, -83.15509841086275 12.930288301759106 C-83.35231966132658 12.726641168489413, -83.54954091179039 12.522994035219723, -83.95781644735568 12.101416024703703 C-84.12023693875304 11.910627673520597, -84.28265743015038 11.719839322337492, -84.70577918442453 11.222816433457206 C-84.96626830393107 10.873784744365425, -85.2267574234376 10.524753055273646, -85.3959130827452 10.29809988219906 C-85.52510262583941 10.099630051756654, -85.65429216893362 9.901160221314246, -86.02538223409623 9.33106622958946 C-86.18889481886426 9.0407330778892, -86.35240740363227 8.75039992618894, -86.591600014714 8.325689224335036 C-86.7347818431034 8.028369094084093, -86.87796367149281 7.731048963833149, -87.09223971428463 7.286100176203095 C-87.24656252046898 6.9049198612992475, -87.40088532665334 6.5237395463954, -87.52524409689568 6.2165709795835475 C-87.6147666223685 5.946943396336804, -87.70428914784134 5.677315813090061, -87.8888338546599 5.121496559358584 C-87.99301668618801 4.724202269617039, -88.09719951771613 4.326907979875495, -88.18151491927281 4.005376811213663 C-88.24913706352606 3.6581513924184677, -88.31675920777933 3.310925973623272, -88.4020846014601 2.8727981106008347 C-88.45961075963345 2.42663658763316, -88.5171369178068 1.9804750646654856, -88.54963653308556 1.7284144663382852 C-88.56540485696964 1.4828100938819277, -88.58117318085371 1.2372057214255703, -88.62356439161238 0.5769283962897952 C-88.62356439161238 0.23782827733294232, -88.62356439161238 -0.10127184162391056, -88.62356439161238 -0.5769283962897864 C-88.5997811558408 -0.9473714909144717, -88.57599792006923 -1.317814585539157, -88.54963653308556 -1.7284144663382603 C-88.50364415592404 -2.0851222613809073, -88.45765177876252 -2.4418300564235547, -88.4020846014601 -2.8727981106008262 C-88.34748346049113 -3.153163429521263, -88.29288231952216 -3.4335287484416996, -88.18151491927283 -4.005376811213655 C-88.12061714028559 -4.237606435756369, -88.05971936129835 -4.469836060299084, -87.8888338546599 -5.121496559358575 C-87.7786926478341 -5.4532243256996775, -87.66855144100832 -5.78495209204078, -87.52524409689569 -6.2165709795835244 C-87.41941475406652 -6.477971502442723, -87.31358541123736 -6.739372025301922, -87.09223971428463 -7.286100176203086 C-86.96718147986242 -7.545786275798372, -86.8421232454402 -7.8054723753936575, -86.591600014714 -8.32568922433503 C-86.41511643520458 -8.639053690654299, -86.23863285569514 -8.95241815697357, -86.02538223409623 -9.331066229589455 C-85.88171447026765 -9.551778495026468, -85.73804670643908 -9.772490760463484, -85.39591308274521 -10.298099882199054 C-85.23893444796049 -10.508436951762748, -85.08195581317577 -10.718774021326443, -84.70577918442454 -11.2228164334572 C-84.46516677625488 -11.505453459327162, -84.22455436808522 -11.788090485197124, -83.95781644735568 -12.101416024703695 C-83.77673717767993 -12.288395237528427, -83.5956579080042 -12.475374450353158, -83.15509841086276 -12.930288301759086 C-82.88355559992861 -13.176896298020413, -82.61201278899446 -13.423504294281738, -82.3009236155402 -13.706027250644407 C-82.07669920673052 -13.884840249808018, -81.85247479792085 -14.06365324897163, -81.3988020488419 -14.425445193623215 C-81.14076342339195 -14.605441768425042, -80.88272479794199 -14.785438343226868, -80.45244072178988 -15.085585888053124 C-80.09956082728286 -15.299503753855728, -79.74668093277583 -15.513421619658331, -79.46572843607089 -15.683736674221008 C-79.06118082764267 -15.894788902412895, -78.65663321921447 -16.105841130604784, -78.44271980411605 -16.217439622243543 C-78.03382101384368 -16.398446987145473, -77.6249222235713 -16.5794543520474, -77.38761858782874 -16.684501632228386 C-77.11776503588128 -16.783810233843944, -76.84791148393381 -16.8831188354595, -76.30476042442518 -17.08300344619203 C-75.8670840306043 -17.212903502467796, -75.42940763678342 -17.342803558743558, -75.19859501037115 -17.411307534702523 C-74.7578561982437 -17.51190329248108, -74.31711738611625 -17.61249905025964, -74.0736678166247 -17.668064825839174 C-73.78145374181156 -17.71530773130097, -73.4892396669984 -17.762550636762764, -72.93460141032112 -17.85222024881843 C-72.545937127323 -17.88971422476207, -72.15727284432486 -17.92720820070571, -71.78607645965283 -17.963017069506055 C-71.3581246182317 -17.976740652102702, -70.93017277681056 -17.990464234699353, -70.6328125 -18 C-70.6328125 -18, -70.6328125 -18, -70.6328125 -18" stroke="none" stroke-width="0" fill="#eee" style=""/><path d="M-70.6328125 -18 C-35.71672128535664 -18, -0.8006300707132823 -18, 70.6328125 -18 M-70.6328125 -18 C-25.41650304958366 -18, 19.79980640083268 -18, 70.6328125 -18 M70.6328125 -18 C70.6328125 -18, 70.6328125 -18, 70.6328125 -18 M70.6328125 -18 C70.6328125 -18, 70.6328125 -18, 70.6328125 -18 M70.6328125 -18 C71.04640152499235 -17.986737004970696, 71.45999054998471 -17.97347400994139, 71.78607645965283 -17.963017069506055 M70.6328125 -18 C70.92071944687353 -17.990767384590644, 71.20862639374708 -17.981534769181284, 71.78607645965283 -17.963017069506055 M71.78607645965283 -17.963017069506055 C72.17695734390266 -17.925309260705514, 72.5678382281525 -17.887601451904974, 72.93460141032111 -17.85222024881843 M71.78607645965283 -17.963017069506055 C72.19536406400792 -17.92353358655428, 72.604651668363 -17.884050103602508, 72.93460141032111 -17.85222024881843 M72.93460141032111 -17.85222024881843 C73.33443286738726 -17.78757859957923, 73.73426432445343 -17.722936950340035, 74.07366781662469 -17.668064825839178 M72.93460141032111 -17.85222024881843 C73.18199373449794 -17.812223776401648, 73.42938605867475 -17.772227303984867, 74.07366781662469 -17.668064825839178 M74.07366781662469 -17.668064825839178 C74.44464015135974 -17.583392811256626, 74.81561248609479 -17.49872079667407, 75.19859501037112 -17.41130753470253 M74.07366781662469 -17.668064825839178 C74.36942853474109 -17.600559371948165, 74.66518925285749 -17.533053918057153, 75.19859501037112 -17.41130753470253 M75.19859501037112 -17.41130753470253 C75.5771114981848 -17.29896582799302, 75.95562798599849 -17.186624121283504, 76.30476042442517 -17.083003446192034 M75.19859501037112 -17.41130753470253 C75.51940339278671 -17.316093290759145, 75.8402117752023 -17.220879046815757, 76.30476042442517 -17.083003446192034 M76.30476042442517 -17.083003446192034 C76.65727206539552 -16.953275903733058, 77.00978370636587 -16.823548361274085, 77.38761858782874 -16.68450163222839 M76.30476042442517 -17.083003446192034 C76.65765962115957 -16.953133279629363, 77.01055881789395 -16.823263113066695, 77.38761858782874 -16.68450163222839 M77.38761858782874 -16.68450163222839 C77.64194468026511 -16.57191901542229, 77.89627077270148 -16.45933639861619, 78.44271980411605 -16.217439622243546 M77.38761858782874 -16.68450163222839 C77.61397824602408 -16.584298926266, 77.84033790421941 -16.484096220303613, 78.44271980411605 -16.217439622243546 M78.44271980411605 -16.217439622243546 C78.73643243187215 -16.064209932475105, 79.03014505962824 -15.910980242706666, 79.46572843607088 -15.683736674221011 M78.44271980411605 -16.217439622243546 C78.77700152213706 -16.04304506406965, 79.11128324015806 -15.868650505895749, 79.46572843607088 -15.683736674221011 M79.46572843607088 -15.683736674221011 C79.68948267999968 -15.548095536625473, 79.9132369239285 -15.412454399029935, 80.45244072178987 -15.085585888053133 M79.46572843607088 -15.683736674221011 C79.77861378815736 -15.494063735067765, 80.09149914024383 -15.30439079591452, 80.45244072178987 -15.085585888053133 M80.45244072178987 -15.085585888053133 C80.811256932921 -14.835291232071492, 81.17007314405213 -14.58499657608985, 81.39880204884189 -14.425445193623219 M80.45244072178987 -15.085585888053133 C80.6742982412401 -14.930827686541882, 80.89615576069032 -14.776069485030629, 81.39880204884189 -14.425445193623219 M81.39880204884189 -14.425445193623219 C81.6196367014523 -14.249335434823326, 81.84047135406271 -14.073225676023432, 82.30092361554019 -13.706027250644421 M81.39880204884189 -14.425445193623219 C81.75712835680332 -14.139689498487492, 82.11545466476475 -13.853933803351765, 82.30092361554019 -13.706027250644421 M82.30092361554019 -13.706027250644421 C82.48015070063873 -13.543257954046735, 82.65937778573728 -13.380488657449051, 83.15509841086275 -12.930288301759097 M82.30092361554019 -13.706027250644421 C82.52809272439818 -13.499718276544847, 82.75526183325616 -13.293409302445275, 83.15509841086275 -12.930288301759097 M83.15509841086275 -12.930288301759097 C83.46066156802247 -12.614769254789932, 83.76622472518218 -12.299250207820767, 83.95781644735568 -12.101416024703706 M83.15509841086275 -12.930288301759097 C83.45260618549446 -12.6230870986093, 83.75011396012617 -12.315885895459504, 83.95781644735568 -12.101416024703706 M83.95781644735568 -12.101416024703706 C84.11690394586732 -11.91454279661682, 84.27599144437897 -11.727669568529933, 84.70577918442453 -11.222816433457204 M83.95781644735568 -12.101416024703706 C84.14962046384463 -11.87611211297807, 84.34142448033357 -11.650808201252435, 84.70577918442453 -11.222816433457204 M84.70577918442453 -11.222816433457204 C84.94944126551411 -10.896331442869405, 85.19310334660369 -10.569846452281608, 85.3959130827452 -10.298099882199057 M84.70577918442453 -11.222816433457204 C84.85131854855193 -11.027806944435326, 84.99685791267932 -10.832797455413447, 85.3959130827452 -10.298099882199057 M85.3959130827452 -10.298099882199057 C85.57003944476858 -10.030595030655993, 85.74416580679197 -9.76309017911293, 86.02538223409623 -9.331066229589453 M85.3959130827452 -10.298099882199057 C85.54732132464413 -10.065496163836581, 85.69872956654307 -9.832892445474103, 86.02538223409623 -9.331066229589453 M86.02538223409623 -9.331066229589453 C86.16031662017825 -9.091476563512593, 86.29525100626029 -8.851886897435733, 86.591600014714 -8.325689224335035 M86.02538223409623 -9.331066229589453 C86.25040032004516 -8.93152383243515, 86.47541840599409 -8.531981435280848, 86.591600014714 -8.325689224335035 M86.591600014714 -8.325689224335035 C86.78545361631049 -7.923148072817332, 86.97930721790696 -7.52060692129963, 87.09223971428463 -7.286100176203093 M86.591600014714 -8.325689224335035 C86.73771204130986 -8.022284474812178, 86.88382406790573 -7.71887972528932, 87.09223971428463 -7.286100176203093 M87.09223971428463 -7.286100176203093 C87.21386555654787 -6.985681994337179, 87.33549139881112 -6.685263812471265, 87.52524409689568 -6.216570979583546 M87.09223971428463 -7.286100176203093 C87.2563804703168 -6.880669335357981, 87.42052122634898 -6.475238494512869, 87.52524409689568 -6.216570979583546 M87.52524409689568 -6.216570979583546 C87.6498508639415 -5.841275250633722, 87.77445763098733 -5.465979521683899, 87.8888338546599 -5.121496559358589 M87.52524409689568 -6.216570979583546 C87.63858983998574 -5.875191660587947, 87.7519355830758 -5.533812341592348, 87.8888338546599 -5.121496559358589 M87.8888338546599 -5.121496559358589 C87.97503719139884 -4.792765875610492, 88.0612405281378 -4.464035191862394, 88.18151491927283 -4.005376811213661 M87.8888338546599 -5.121496559358589 C87.98811388220393 -4.742898786097061, 88.08739390974796 -4.364301012835532, 88.18151491927283 -4.005376811213661 M88.18151491927283 -4.005376811213661 C88.23066142269494 -3.7530199032582865, 88.27980792611706 -3.5006629953029127, 88.4020846014601 -2.872798110600833 M88.18151491927283 -4.005376811213661 C88.2518036747434 -3.6444589074059937, 88.32209243021397 -3.2835410035983266, 88.4020846014601 -2.872798110600833 M88.4020846014601 -2.872798110600833 C88.45533752228822 -2.4597789726365695, 88.50859044311635 -2.046759834672306, 88.54963653308556 -1.728414466338275 M88.4020846014601 -2.872798110600833 C88.4333839154398 -2.630046812122881, 88.4646832294195 -2.387295513644929, 88.54963653308556 -1.728414466338275 M88.54963653308556 -1.728414466338275 C88.56996327001394 -1.4118091312318855, 88.59029000694234 -1.095203796125496, 88.62356439161238 -0.576928396289793 M88.54963653308556 -1.728414466338275 C88.57853805086272 -1.27824999807375, 88.60743956863988 -0.8280855298092252, 88.62356439161238 -0.576928396289793 M88.62356439161238 -0.576928396289793 C88.62356439161238 -0.27834077789320427, 88.62356439161238 0.020246840503384433, 88.62356439161238 0.5769283962897886 M88.62356439161238 -0.576928396289793 C88.62356439161238 -0.15772045779058402, 88.62356439161238 0.26148748070862493, 88.62356439161238 0.5769283962897886 M88.62356439161238 0.5769283962897886 C88.60592427247003 0.8516874907795557, 88.5882841533277 1.1264465852693228, 88.54963653308556 1.7284144663382628 M88.62356439161238 0.5769283962897886 C88.59713476289993 0.9885911970983914, 88.57070513418749 1.4002539979069941, 88.54963653308556 1.7284144663382628 M88.54963653308556 1.7284144663382628 C88.50498702560483 2.07470722931626, 88.4603375181241 2.4209999922942567, 88.4020846014601 2.8727981106008205 M88.54963653308556 1.7284144663382628 C88.4940476264308 2.1595510490813483, 88.43845871977605 2.5906876318244336, 88.4020846014601 2.8727981106008205 M88.4020846014601 2.8727981106008205 C88.357518511664 3.101635562657443, 88.31295242186788 3.330473014714065, 88.18151491927283 4.005376811213657 M88.4020846014601 2.8727981106008205 C88.3371341615617 3.2063048921987787, 88.2721837216633 3.5398116737967373, 88.18151491927283 4.005376811213657 M88.18151491927283 4.005376811213657 C88.0966572042292 4.328976056241318, 88.01179948918559 4.652575301268979, 87.8888338546599 5.121496559358577 M88.18151491927283 4.005376811213657 C88.10555629711983 4.295039958158126, 88.02959767496684 4.584703105102596, 87.8888338546599 5.121496559358577 M87.8888338546599 5.121496559358577 C87.78126847903143 5.445466332536325, 87.67370310340296 5.769436105714075, 87.52524409689569 6.216570979583534 M87.8888338546599 5.121496559358577 C87.77751124025721 5.456782537089554, 87.66618862585453 5.79206851482053, 87.52524409689569 6.216570979583534 M87.52524409689569 6.216570979583534 C87.39112488136719 6.547848034957558, 87.25700566583869 6.879125090331581, 87.09223971428463 7.286100176203088 M87.52524409689569 6.216570979583534 C87.40401785417143 6.51600214286242, 87.28279161144717 6.815433306141305, 87.09223971428463 7.286100176203088 M87.09223971428463 7.286100176203088 C86.98777646353221 7.50302035154597, 86.88331321277981 7.719940526888853, 86.591600014714 8.325689224335031 M87.09223971428463 7.286100176203088 C86.99047271133327 7.497421535296642, 86.88870570838189 7.708742894390196, 86.591600014714 8.325689224335031 M86.591600014714 8.325689224335031 C86.46198081571049 8.555841237532928, 86.33236161670698 8.785993250730826, 86.02538223409623 9.33106622958945 M86.591600014714 8.325689224335031 C86.44160824089205 8.592014807069978, 86.29161646707009 8.858340389804923, 86.02538223409623 9.33106622958945 M86.02538223409623 9.33106622958945 C85.82509273344616 9.63876468784954, 85.62480323279607 9.94646314610963, 85.39591308274521 10.298099882199049 M86.02538223409623 9.33106622958945 C85.7801357035556 9.707830758414278, 85.53488917301497 10.084595287239107, 85.39591308274521 10.298099882199049 M85.39591308274521 10.298099882199049 C85.21487180578366 10.540678700127906, 85.0338305288221 10.783257518056763, 84.70577918442454 11.2228164334572 M85.39591308274521 10.298099882199049 C85.12426913163306 10.662078029416284, 84.85262518052092 11.02605617663352, 84.70577918442454 11.2228164334572 M84.70577918442454 11.2228164334572 C84.43126388437716 11.545277725255934, 84.15674858432978 11.867739017054665, 83.95781644735568 12.101416024703695 M84.70577918442454 11.2228164334572 C84.47468325312235 11.49427486409576, 84.24358732182016 11.765733294734318, 83.95781644735568 12.101416024703695 M83.95781644735568 12.101416024703695 C83.69816811122257 12.369524249285988, 83.43851977508945 12.637632473868281, 83.15509841086276 12.930288301759093 M83.95781644735568 12.101416024703695 C83.78521812540615 12.279637962702362, 83.61261980345661 12.457859900701028, 83.15509841086276 12.930288301759093 M83.15509841086276 12.930288301759093 C82.88630788018038 13.174396749917104, 82.617517349498 13.418505198075113, 82.3009236155402 13.706027250644409 M83.15509841086276 12.930288301759093 C82.95029579944413 13.116284614669336, 82.74549318802549 13.302280927579579, 82.3009236155402 13.706027250644409 M82.3009236155402 13.706027250644409 C82.08751750858205 13.876212941968518, 81.87411140162392 14.046398633292629, 81.39880204884189 14.425445193623217 M82.3009236155402 13.706027250644409 C82.01751764112966 13.932035973487105, 81.73411166671912 14.158044696329801, 81.39880204884189 14.425445193623217 M81.39880204884189 14.425445193623217 C81.10816393511715 14.628181760236426, 80.81752582139242 14.830918326849636, 80.45244072178988 15.085585888053126 M81.39880204884189 14.425445193623217 C81.10395730784924 14.631116121316564, 80.8091125668566 14.836787049009912, 80.45244072178988 15.085585888053126 M80.45244072178988 15.085585888053126 C80.0712689130982 15.316654478496211, 79.69009710440652 15.547723068939296, 79.46572843607088 15.68373667422101 M80.45244072178988 15.085585888053126 C80.15710328098423 15.26462118026496, 79.86176584017859 15.443656472476794, 79.46572843607088 15.68373667422101 M79.46572843607088 15.68373667422101 C79.06638013213102 15.892076428591638, 78.66703182819116 16.100416182962267, 78.44271980411605 16.217439622243543 M79.46572843607088 15.68373667422101 C79.22671909087761 15.808427696515546, 78.98770974568434 15.933118718810084, 78.44271980411605 16.217439622243543 M78.44271980411605 16.217439622243543 C78.18407583113185 16.33193363668279, 77.92543185814765 16.446427651122043, 77.38761858782874 16.68450163222839 M78.44271980411605 16.217439622243543 C78.21805395343246 16.3168925298655, 77.99338810274887 16.41634543748745, 77.38761858782874 16.68450163222839 M77.38761858782874 16.68450163222839 C76.99869447426109 16.82762930119638, 76.60977036069342 16.970756970164373, 76.30476042442518 17.08300344619203 M77.38761858782874 16.68450163222839 C77.07249990875846 16.800468219051393, 76.75738122968818 16.916434805874392, 76.30476042442518 17.08300344619203 M76.30476042442518 17.08300344619203 C75.94909361343386 17.18856348876107, 75.59342680244254 17.294123531330108, 75.19859501037115 17.411307534702527 M76.30476042442518 17.08300344619203 C75.9133574001309 17.199169809286172, 75.52195437583664 17.315336172380313, 75.19859501037115 17.411307534702527 M75.19859501037115 17.411307534702527 C74.7515491436421 17.513342836536523, 74.30450327691302 17.615378138370517, 74.0736678166247 17.668064825839174 M75.19859501037115 17.411307534702527 C74.87614175210355 17.48490538669795, 74.55368849383595 17.55850323869337, 74.0736678166247 17.668064825839174 M74.0736678166247 17.668064825839174 C73.7884364715084 17.71417881771546, 73.50320512639212 17.760292809591746, 72.93460141032112 17.85222024881843 M74.0736678166247 17.668064825839174 C73.75340777873035 17.719841985112247, 73.43314774083599 17.77161914438532, 72.93460141032112 17.85222024881843 M72.93460141032112 17.85222024881843 C72.58903165116241 17.88555694787603, 72.24346189200371 17.91889364693363, 71.78607645965283 17.963017069506055 M72.93460141032112 17.85222024881843 C72.70160519375086 17.874697113612015, 72.4686089771806 17.8971739784056, 71.78607645965283 17.963017069506055 M71.78607645965283 17.963017069506055 C71.46557336012647 17.973294980086425, 71.1450702606001 17.9835728906668, 70.6328125 18 M71.78607645965283 17.963017069506055 C71.54201195762379 17.970843743043392, 71.29794745559474 17.97867041658073, 70.6328125 18 M70.6328125 18 C70.6328125 18, 70.6328125 18, 70.6328125 18 M70.6328125 18 C70.6328125 18, 70.6328125 18, 70.6328125 18 M70.6328125 18 C36.737657694899184 18, 2.842502889798368 18, -70.6328125 18 M70.6328125 18 C40.47642943300484 18, 10.32004636600967 18, -70.6328125 18 M-70.6328125 18 C-70.91468836002116 17.9909607898072, -71.19656422004232 17.981921579614397, -71.78607645965282 17.963017069506055 M-70.6328125 18 C-71.06970433197546 17.98598972930691, -71.50659616395092 17.97197945861382, -71.78607645965282 17.963017069506055 M-71.78607645965282 17.963017069506055 C-72.23502086283298 17.919707942865905, -72.68396526601313 17.876398816225755, -72.93460141032111 17.85222024881843 M-71.78607645965282 17.963017069506055 C-72.09752570270048 17.932971935601763, -72.40897494574814 17.902926801697472, -72.93460141032111 17.85222024881843 M-72.93460141032111 17.85222024881843 C-73.36781599802713 17.782181473907425, -73.80103058573314 17.71214269899642, -74.0736678166247 17.668064825839174 M-72.93460141032111 17.85222024881843 C-73.35547369609137 17.784176881568417, -73.77634598186162 17.716133514318404, -74.0736678166247 17.668064825839174 M-74.0736678166247 17.668064825839174 C-74.4816095716333 17.574954782327197, -74.8895513266419 17.481844738815223, -75.19859501037112 17.41130753470253 M-74.0736678166247 17.668064825839174 C-74.46104916899242 17.579647560060813, -74.84843052136013 17.491230294282452, -75.19859501037112 17.41130753470253 M-75.19859501037112 17.41130753470253 C-75.43059264073472 17.34245185579103, -75.6625902710983 17.27359617687953, -76.30476042442517 17.083003446192038 M-75.19859501037112 17.41130753470253 C-75.55806009958222 17.304620183095434, -75.91752518879333 17.197932831488338, -76.30476042442517 17.083003446192038 M-76.30476042442517 17.083003446192038 C-76.71682536675563 16.931359738327856, -77.12889030908609 16.779716030463675, -77.38761858782874 16.68450163222839 M-76.30476042442517 17.083003446192038 C-76.70003680432836 16.93753808449286, -77.09531318423156 16.792072722793687, -77.38761858782874 16.68450163222839 M-77.38761858782874 16.68450163222839 C-77.61660337986977 16.583136857384336, -77.84558817191082 16.481772082540285, -78.44271980411604 16.217439622243546 M-77.38761858782874 16.68450163222839 C-77.80132520819838 16.501365983561612, -78.21503182856802 16.318230334894835, -78.44271980411604 16.217439622243546 M-78.44271980411604 16.217439622243546 C-78.6954704124785 16.085579791755343, -78.94822102084096 15.953719961267138, -79.46572843607088 15.683736674221013 M-78.44271980411604 16.217439622243546 C-78.75724190787288 16.053353642174624, -79.07176401162972 15.8892676621057, -79.46572843607088 15.683736674221013 M-79.46572843607088 15.683736674221013 C-79.70370824341713 15.539471916483137, -79.94168805076336 15.395207158745263, -80.45244072178987 15.085585888053139 M-79.46572843607088 15.683736674221013 C-79.76115361780356 15.50464819294359, -80.05657879953624 15.325559711666168, -80.45244072178987 15.085585888053139 M-80.45244072178987 15.085585888053139 C-80.709208025903 14.906476131981265, -80.96597533001614 14.727366375909392, -81.39880204884189 14.42544519362322 M-80.45244072178987 15.085585888053139 C-80.78431697896279 14.854083377740928, -81.11619323613571 14.622580867428718, -81.39880204884189 14.42544519362322 M-81.39880204884189 14.42544519362322 C-81.7243728873568 14.165811113711445, -82.04994372587171 13.906177033799672, -82.30092361554019 13.706027250644423 M-81.39880204884189 14.42544519362322 C-81.60281538351171 14.262749988246874, -81.80682871818154 14.100054782870526, -82.30092361554019 13.706027250644423 M-82.30092361554019 13.706027250644423 C-82.5631771901027 13.467855493076089, -82.82543076466523 13.229683735507752, -83.15509841086275 12.930288301759106 M-82.30092361554019 13.706027250644423 C-82.60366769041156 13.431083092015525, -82.90641176528293 13.156138933386627, -83.15509841086275 12.930288301759106 M-83.15509841086275 12.930288301759106 C-83.41116851828656 12.665874892097214, -83.66723862571037 12.401461482435325, -83.95781644735568 12.101416024703703 M-83.15509841086275 12.930288301759106 C-83.40460360009014 12.672653709124823, -83.65410878931753 12.415019116490543, -83.95781644735568 12.101416024703703 M-83.95781644735568 12.101416024703703 C-84.17581547528467 11.845341962002223, -84.39381450321365 11.589267899300744, -84.70577918442453 11.222816433457206 M-83.95781644735568 12.101416024703703 C-84.18023611865326 11.840149222759752, -84.40265578995081 11.5788824208158, -84.70577918442453 11.222816433457206 M-84.70577918442453 11.222816433457206 C-84.9110073544182 10.947829382336604, -85.11623552441186 10.672842331216001, -85.3959130827452 10.29809988219906 M-84.70577918442453 11.222816433457206 C-84.90463679402963 10.956365352887296, -85.10349440363473 10.689914272317388, -85.3959130827452 10.29809988219906 M-85.3959130827452 10.29809988219906 C-85.6326330764385 9.934434403667455, -85.86935307013181 9.57076892513585, -86.02538223409623 9.33106622958946 M-85.3959130827452 10.29809988219906 C-85.57702820729182 10.019858414674808, -85.75814333183843 9.741616947150556, -86.02538223409623 9.33106622958946 M-86.02538223409623 9.33106622958946 C-86.22870719051991 8.970042180574794, -86.4320321469436 8.609018131560127, -86.591600014714 8.325689224335036 M-86.02538223409623 9.33106622958946 C-86.19112575671508 9.0367718218281, -86.35686927933394 8.742477414066741, -86.591600014714 8.325689224335036 M-86.591600014714 8.325689224335036 C-86.76081309431403 7.9743146439657, -86.93002617391404 7.622940063596364, -87.09223971428463 7.286100176203095 M-86.591600014714 8.325689224335036 C-86.78881516615529 7.916167742638955, -86.98603031759657 7.506646260942874, -87.09223971428463 7.286100176203095 M-87.09223971428463 7.286100176203095 C-87.23886869806054 6.923923751398019, -87.38549768183647 6.561747326592943, -87.52524409689568 6.2165709795835475 M-87.09223971428463 7.286100176203095 C-87.21603250281095 6.980329595172332, -87.33982529133726 6.67455901414157, -87.52524409689568 6.2165709795835475 M-87.52524409689568 6.2165709795835475 C-87.6655402648862 5.7940212747548205, -87.80583643287675 5.371471569926093, -87.8888338546599 5.121496559358584 M-87.52524409689568 6.2165709795835475 C-87.62085265468026 5.928613236691353, -87.71646121246485 5.640655493799159, -87.8888338546599 5.121496559358584 M-87.8888338546599 5.121496559358584 C-87.99756730346698 4.706848793682356, -88.10630075227405 4.292201028006128, -88.18151491927281 4.005376811213663 M-87.8888338546599 5.121496559358584 C-87.98500823785656 4.754741953988345, -88.08118262105322 4.387987348618106, -88.18151491927281 4.005376811213663 M-88.18151491927281 4.005376811213663 C-88.25514884546126 3.6272821627717806, -88.32878277164973 3.249187514329898, -88.4020846014601 2.8727981106008347 M-88.18151491927281 4.005376811213663 C-88.24016995206145 3.7041956145895774, -88.2988249848501 3.4030144179654913, -88.4020846014601 2.8727981106008347 M-88.4020846014601 2.8727981106008347 C-88.44356725188588 2.551066865574317, -88.48504990231167 2.2293356205477997, -88.54963653308556 1.7284144663382852 M-88.4020846014601 2.8727981106008347 C-88.43622970649719 2.607975418646845, -88.47037481153427 2.3431527266928547, -88.54963653308556 1.7284144663382852 M-88.54963653308556 1.7284144663382852 C-88.56469732776584 1.4938304323382954, -88.57975812244612 1.2592463983383055, -88.62356439161238 0.5769283962897952 M-88.54963653308556 1.7284144663382852 C-88.57778376081768 1.289998674377284, -88.60593098854979 0.8515828824162828, -88.62356439161238 0.5769283962897952 M-88.62356439161238 0.5769283962897952 C-88.62356439161238 0.3181701436670918, -88.62356439161238 0.059411891044388376, -88.62356439161238 -0.5769283962897864 M-88.62356439161238 0.5769283962897952 C-88.62356439161238 0.2721902346059617, -88.62356439161238 -0.0325479270778718, -88.62356439161238 -0.5769283962897864 M-88.62356439161238 -0.5769283962897864 C-88.60405474694014 -0.8808068599616462, -88.5845451022679 -1.184685323633506, -88.54963653308556 -1.7284144663382603 M-88.62356439161238 -0.5769283962897864 C-88.60733329373143 -0.829740847292013, -88.59110219585047 -1.0825532982942396, -88.54963653308556 -1.7284144663382603 M-88.54963653308556 -1.7284144663382603 C-88.4994929723429 -2.1173180214922334, -88.44934941160025 -2.5062215766462064, -88.4020846014601 -2.8727981106008262 M-88.54963653308556 -1.7284144663382603 C-88.49273311392126 -2.1697461485255616, -88.43582969475695 -2.611077830712863, -88.4020846014601 -2.8727981106008262 M-88.4020846014601 -2.8727981106008262 C-88.34259525696389 -3.1782633215114835, -88.28310591246768 -3.4837285324221403, -88.18151491927283 -4.005376811213655 M-88.4020846014601 -2.8727981106008262 C-88.33906053104866 -3.196413392021591, -88.27603646063721 -3.520028673442355, -88.18151491927283 -4.005376811213655 M-88.18151491927283 -4.005376811213655 C-88.07615014352426 -4.407178366525616, -87.97078536777569 -4.808979921837578, -87.8888338546599 -5.121496559358575 M-88.18151491927283 -4.005376811213655 C-88.07155740660255 -4.424692462912883, -87.96159989393226 -4.844008114612111, -87.8888338546599 -5.121496559358575 M-87.8888338546599 -5.121496559358575 C-87.80906441304377 -5.361749407971805, -87.72929497142765 -5.602002256585035, -87.52524409689569 -6.2165709795835244 M-87.8888338546599 -5.121496559358575 C-87.77848957369338 -5.453835952656101, -87.66814529272685 -5.7861753459536285, -87.52524409689569 -6.2165709795835244 M-87.52524409689569 -6.2165709795835244 C-87.36099705346119 -6.622264352367301, -87.19675001002669 -7.027957725151078, -87.09223971428463 -7.286100176203086 M-87.52524409689569 -6.2165709795835244 C-87.4223816634062 -6.470643179012437, -87.31951922991671 -6.724715378441348, -87.09223971428463 -7.286100176203086 M-87.09223971428463 -7.286100176203086 C-86.93717421337647 -7.608097006709084, -86.78210871246831 -7.9300938372150815, -86.591600014714 -8.32568922433503 M-87.09223971428463 -7.286100176203086 C-86.97972906087513 -7.51973095535024, -86.86721840746564 -7.7533617344973935, -86.591600014714 -8.32568922433503 M-86.591600014714 -8.32568922433503 C-86.4558261655223 -8.566769442167974, -86.3200523163306 -8.807849660000919, -86.02538223409623 -9.331066229589455 M-86.591600014714 -8.32568922433503 C-86.40357796748773 -8.659541408482106, -86.21555592026147 -8.993393592629182, -86.02538223409623 -9.331066229589455 M-86.02538223409623 -9.331066229589455 C-85.77611280284775 -9.714011014201649, -85.52684337159927 -10.096955798813841, -85.39591308274521 -10.298099882199054 M-86.02538223409623 -9.331066229589455 C-85.81075445661742 -9.660792131424902, -85.59612667913862 -9.99051803326035, -85.39591308274521 -10.298099882199054 M-85.39591308274521 -10.298099882199054 C-85.1858150045776 -10.579612166614961, -84.97571692641002 -10.861124451030866, -84.70577918442454 -11.2228164334572 M-85.39591308274521 -10.298099882199054 C-85.23448266981659 -10.514401928910239, -85.07305225688798 -10.730703975621426, -84.70577918442454 -11.2228164334572 M-84.70577918442454 -11.2228164334572 C-84.52651520030909 -11.433390109608876, -84.34725121619363 -11.64396378576055, -83.95781644735568 -12.101416024703695 M-84.70577918442454 -11.2228164334572 C-84.48549950897599 -11.481569474507685, -84.26521983352744 -11.740322515558168, -83.95781644735568 -12.101416024703695 M-83.95781644735568 -12.101416024703695 C-83.70911977533783 -12.35821575693864, -83.46042310331997 -12.615015489173581, -83.15509841086276 -12.930288301759086 M-83.95781644735568 -12.101416024703695 C-83.68252508419607 -12.385676959007249, -83.40723372103646 -12.6699378933108, -83.15509841086276 -12.930288301759086 M-83.15509841086276 -12.930288301759086 C-82.87674470642197 -13.183081771296777, -82.59839100198117 -13.435875240834466, -82.3009236155402 -13.706027250644407 M-83.15509841086276 -12.930288301759086 C-82.94540978448258 -13.12072196418054, -82.7357211581024 -13.311155626601993, -82.3009236155402 -13.706027250644407 M-82.3009236155402 -13.706027250644407 C-82.1092652141475 -13.858869725510836, -81.91760681275478 -14.011712200377264, -81.3988020488419 -14.425445193623215 M-82.3009236155402 -13.706027250644407 C-82.0780852991851 -13.883734877960958, -81.85524698283001 -14.061442505277508, -81.3988020488419 -14.425445193623215 M-81.3988020488419 -14.425445193623215 C-81.08770267925394 -14.642454646573125, -80.77660330966596 -14.859464099523034, -80.45244072178988 -15.085585888053124 M-81.3988020488419 -14.425445193623215 C-81.15118543115057 -14.598171826974069, -80.90356881345923 -14.770898460324922, -80.45244072178988 -15.085585888053124 M-80.45244072178988 -15.085585888053124 C-80.13011532619618 -15.280981437056308, -79.80778993060247 -15.476376986059494, -79.46572843607089 -15.683736674221008 M-80.45244072178988 -15.085585888053124 C-80.16868327523144 -15.2576013197007, -79.88492582867302 -15.429616751348277, -79.46572843607089 -15.683736674221008 M-79.46572843607089 -15.683736674221008 C-79.07383023570418 -15.888189714329544, -78.68193203533747 -16.092642754438078, -78.44271980411605 -16.217439622243543 M-79.46572843607089 -15.683736674221008 C-79.08444760257551 -15.88265064083352, -78.70316676908013 -16.08156460744603, -78.44271980411605 -16.217439622243543 M-78.44271980411605 -16.217439622243543 C-78.02345386001593 -16.40303621861432, -77.60418791591582 -16.588632814985104, -77.38761858782874 -16.684501632228386 M-78.44271980411605 -16.217439622243543 C-78.16955241038879 -16.338362724223774, -77.89638501666153 -16.459285826204006, -77.38761858782874 -16.684501632228386 M-77.38761858782874 -16.684501632228386 C-77.04942709296654 -16.808959228545586, -76.71123559810432 -16.93341682486279, -76.30476042442518 -17.08300344619203 M-77.38761858782874 -16.684501632228386 C-77.14124663887777 -16.775168786879675, -76.8948746899268 -16.865835941530964, -76.30476042442518 -17.08300344619203 M-76.30476042442518 -17.08300344619203 C-75.989500269664 -17.176571005290825, -75.67424011490283 -17.270138564389622, -75.19859501037115 -17.411307534702523 M-76.30476042442518 -17.08300344619203 C-75.8967531319908 -17.20409786891257, -75.48874583955644 -17.325192291633112, -75.19859501037115 -17.411307534702523 M-75.19859501037115 -17.411307534702523 C-74.86758122755676 -17.48685927056311, -74.53656744474235 -17.562411006423694, -74.0736678166247 -17.668064825839174 M-75.19859501037115 -17.411307534702523 C-74.96700010673267 -17.46416756016002, -74.7354052030942 -17.51702758561752, -74.0736678166247 -17.668064825839174 M-74.0736678166247 -17.668064825839174 C-73.67118434733239 -17.733135231848436, -73.26870087804008 -17.7982056378577, -72.93460141032112 -17.85222024881843 M-74.0736678166247 -17.668064825839174 C-73.74381440006954 -17.721392968176367, -73.41396098351436 -17.774721110513557, -72.93460141032112 -17.85222024881843 M-72.93460141032112 -17.85222024881843 C-72.56468897914542 -17.887905256169162, -72.19477654796971 -17.923590263519895, -71.78607645965283 -17.963017069506055 M-72.93460141032112 -17.85222024881843 C-72.52604381153775 -17.89163330901066, -72.11748621275437 -17.931046369202893, -71.78607645965283 -17.963017069506055 M-71.78607645965283 -17.963017069506055 C-71.50857572177647 -17.97191597804733, -71.2310749839001 -17.980814886588607, -70.6328125 -18 M-71.78607645965283 -17.963017069506055 C-71.45437007266978 -17.973654247709288, -71.12266368568672 -17.98429142591252, -70.6328125 -18 M-70.6328125 -18 C-70.6328125 -18, -70.6328125 -18, -70.6328125 -18 M-70.6328125 -18 C-70.6328125 -18, -70.6328125 -18, -70.6328125 -18" stroke="#999" stroke-width="1.3" fill="none" stroke-dasharray="0 0" style=""/></g><g class="label" style="" transform="translate(-76.6328125, -10.5)"><rect/><foreignObject width="153.265625" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>fulfilled — nothing to do</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-P-5" data-look="classic" transform="translate(363.09088134765625, 478.609375)"><rect class="basic label-container" style="" x="-127.84375" y="-46.5" width="255.6875" height="93"/><g class="label" style="" transform="translate(-97.84375, -31.5)"><rect/><foreignObject width="195.6875" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>make drift-plan<br />what changed, what to review,<br />previous rationale</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-REV-7" data-look="classic" transform="translate(251.76666259765625, 621.609375)"><rect class="basic label-container" style="" x="-102.1484375" y="-46.5" width="204.296875" height="93"/><g class="label" style="" transform="translate(-72.1484375, -31.5)"><rect/><foreignObject width="144.296875" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>review the target docs<br />against the change;<br />fix what is stale</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-ADD-9" data-look="classic" transform="translate(251.76666259765625, 764.609375)"><rect class="basic label-container" style="" x="-109.3828125" y="-46.5" width="218.765625" height="93"/><g class="label" style="" transform="translate(-79.3828125, -31.5)"><rect/><foreignObject width="158.765625" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>git add the trigger files<br />(digest reads the INDEX,<br />not the working tree)</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-ACK-11" data-look="classic" transform="translate(251.76666259765625, 928.609375)"><rect class="basic label-container" style="" x="-130" y="-67.5" width="260" height="135"/><g class="label" style="" transform="translate(-100, -52.5)"><rect/><foreignObject width="200" height="105"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table; white-space: break-spaces; line-height: 1.5; max-width: 200px; text-align: center; width: 200px;"><span class="nodeLabel"><p>make drift-ack CONTRACT=... RATIONALE=...<br />runs verify_commands, recomputes digest,<br />writes + STAGES the ack file</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-COMMIT-13" data-look="classic" transform="translate(251.76666259765625, 1082.109375)"><rect class="basic label-container" style="" x="-113.5859375" y="-36" width="227.171875" height="72"/><g class="label" style="" transform="translate(-83.5859375, -21)"><rect/><foreignObject width="167.171875" height="42"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>one commit: code change<br />+ doc fix + ack together</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-PRV-15" data-look="classic" transform="translate(251.76666259765625, 1214.609375)"><rect class="basic label-container" style="" x="-99.3828125" y="-46.5" width="198.765625" height="93"/><g class="label" style="" transform="translate(-69.3828125, -31.5)"><rect/><foreignObject width="138.765625" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>PR: reviewer sees the<br />rationale in the diff;<br />CI lint-drift re-checks</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-MERGE-17" data-look="classic" transform="translate(363.09088134765625, 1419.3671875)"><polygon points="108.2578125,0 216.515625,-108.2578125 108.2578125,-216.515625 0,-108.2578125" class="label-container" transform="translate(-107.7578125, 108.2578125)"/><g class="label" style="" transform="translate(-61.7578125, -31.5)"><rect/><foreignObject width="123.515625" height="63"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>after merge:<br />digest recomputed<br />on the merged tree</p></span></div></foreignObject></g></g><g class="node default" id="dg3-flowchart-DONE-19" data-look="classic" transform="translate(363.09088134765625, 1616.625)"><g class="basic label-container outer-path"><path d="M-17.8203125 -18 C-5.110342297759384 -18, 7.599627904481231 -18, 17.8203125 -18 C17.8203125 -18, 17.8203125 -18, 17.8203125 -18 C18.122160877145593 -17.990320310057164, 18.42400925429118 -17.98064062011433, 18.973576459652833 -17.963017069506055 C19.29843529327726 -17.931678328526214, 19.62329412690169 -17.900339587546373, 20.122101410321108 -17.85222024881843 C20.43275759493655 -17.801995766064497, 20.743413779551993 -17.751771283310568, 21.2611678166247 -17.668064825839178 C21.62543493205772 -17.584923233806666, 21.989702047490738 -17.501781641774155, 22.38609501037113 -17.41130753470253 C22.66803891321196 -17.32762806429312, 22.949982816052792 -17.243948593883704, 23.492260424425172 -17.083003446192034 C23.8806864263557 -16.940059086911, 24.269112428286228 -16.79711472762996, 24.575118587828733 -16.68450163222839 C24.888784117881027 -16.545651206631597, 25.202449647933324 -16.406800781034807, 25.630219804116045 -16.217439622243546 C25.98278979402665 -16.03350408433904, 26.33535978393726 -15.849568546434531, 26.653228436070876 -15.683736674221011 C26.880586784682784 -15.545910707186511, 27.107945133294688 -15.408084740152013, 27.639940721789877 -15.085585888053133 C27.87406088974348 -14.922273793787424, 28.10818105769708 -14.758961699521716, 28.586302048841887 -14.425445193623219 C28.785959772752154 -14.26622347191995, 28.985617496662417 -14.107001750216684, 29.488423615540185 -13.706027250644421 C29.726047029059814 -13.49022395365689, 29.963670442579442 -13.27442065666936, 30.342598410862756 -12.930288301759097 C30.5619501785132 -12.703789592544963, 30.781301946163648 -12.47729088333083, 31.145316447355675 -12.101416024703706 C31.35254703689277 -11.857991182478846, 31.55977762642986 -11.614566340253988, 31.893279184424536 -11.222816433457204 C32.07186474361883 -10.983528053615983, 32.250450302813135 -10.74423967377476, 32.5834130827452 -10.298099882199057 C32.72690947173561 -10.07765089453405, 32.870405860726024 -9.857201906869046, 33.21288223409623 -9.331066229589453 C33.41832018591601 -8.966290343285023, 33.6237581377358 -8.601514456980592, 33.779100014714004 -8.325689224335035 C33.9514441752706 -7.9678128873649525, 34.12378833582721 -7.60993655039487, 34.279739714284624 -7.286100176203093 C34.40279698192898 -6.982146345971117, 34.525854249573335 -6.678192515739141, 34.71274409689568 -6.216570979583546 C34.82918624024479 -5.865865796352101, 34.9456283835939 -5.515160613120656, 35.07633385465989 -5.121496559358589 C35.154772483472854 -4.822376072177986, 35.23321111228582 -4.523255584997382, 35.36901491927283 -4.005376811213661 C35.442730803935454 -3.6268613233288325, 35.51644668859807 -3.2483458354440042, 35.5895846014601 -2.872798110600833 C35.63917992777905 -2.488146553376045, 35.68877525409798 -2.103494996151257, 35.73713653308557 -1.728414466338275 C35.75558968676669 -1.440991702637222, 35.77404284044781 -1.1535689389361692, 35.81106439161238 -0.576928396289793 C35.81106439161238 -0.30410240247221987, 35.81106439161238 -0.03127640865464676, 35.81106439161238 0.5769283962897886 C35.790414199552266 0.898571806161164, 35.76976400749214 1.2202152160325395, 35.73713653308557 1.7284144663382628 C35.68176659108664 2.157852802450149, 35.626396649087724 2.587291138562035, 35.5895846014601 2.8727981106008205 C35.51211445587198 3.2705909348747815, 35.43464431028385 3.6683837591487425, 35.36901491927283 4.005376811213657 C35.30420546446209 4.252523351002377, 35.23939600965136 4.499669890791098, 35.07633385465989 5.121496559358577 C34.931782875542595 5.5568610773890486, 34.78723189642529 5.992225595419519, 34.71274409689569 6.216570979583534 C34.555579341156665 6.604770964832762, 34.39841458541764 6.9929709500819905, 34.279739714284624 7.286100176203088 C34.16661128937681 7.521013771084684, 34.053482864469004 7.755927365966279, 33.779100014714004 8.325689224335031 C33.66359277029779 8.530784033172822, 33.54808552588158 8.73587884201061, 33.21288223409623 9.33106622958945 C33.05721085727357 9.570219268006788, 32.901539480450914 9.809372306424128, 32.58341308274521 10.298099882199049 C32.421304303409464 10.515310878095338, 32.25919552407372 10.732521873991628, 31.89327918442454 11.2228164334572 C31.63273723909467 11.528863827698332, 31.372195293764804 11.834911221939464, 31.145316447355682 12.101416024703695 C30.831579367855923 12.425375317580464, 30.517842288356164 12.749334610457234, 30.34259841086276 12.930288301759093 C30.051938231866803 13.194258185876308, 29.761278052870846 13.458228069993524, 29.488423615540203 13.706027250644409 C29.172406120820227 13.958042793104878, 28.856388626100248 14.210058335565346, 28.586302048841894 14.425445193623217 C28.25959979225533 14.653338545105257, 27.932897535668765 14.881231896587297, 27.639940721789884 15.085585888053126 C27.288923042046875 15.298374868322194, 26.93790536230387 15.51116384859126, 26.653228436070883 15.68373667422101 C26.32952367725944 15.852613239550145, 26.005818918447996 16.02148980487928, 25.63021980411605 16.217439622243543 C25.413765228278983 16.313257642069754, 25.197310652441917 16.409075661895965, 24.57511858782874 16.68450163222839 C24.171122289245588 16.833176004607836, 23.767125990662436 16.981850376987282, 23.492260424425183 17.08300344619203 C23.270041252880336 17.148956929854307, 23.04782208133549 17.214910413516584, 22.38609501037115 17.411307534702527 C21.983578313141262 17.503179344178324, 21.581061615911374 17.595051153654122, 21.261167816624706 17.668064825839174 C20.963262472520217 17.71622785158354, 20.665357128415724 17.76439087732791, 20.122101410321118 17.85222024881843 C19.845722208820945 17.87888221799253, 19.569343007320768 17.905544187166633, 18.973576459652833 17.963017069506055 C18.656322454217793 17.973190787962896, 18.339068448782754 17.983364506419736, 17.820312500000004 18 C17.820312500000004 18, 17.8203125 18, 17.8203125 18 C4.04888110928202 18, -9.72255028143596 18, -17.820312499999996 18 C-18.221651639527895 17.987129834954587, -18.62299077905579 17.97425966990917, -18.973576459652826 17.963017069506055 C-19.303117990572847 17.931226594349877, -19.63265952149287 17.899436119193698, -20.12210141032111 17.85222024881843 C-20.577510014809175 17.778593317391273, -21.032918619297238 17.704966385964116, -21.2611678166247 17.668064825839174 C-21.619293997448917 17.58632486205778, -21.97742017827314 17.504584898276384, -22.38609501037113 17.41130753470253 C-22.774688876529947 17.29597491499742, -23.163282742688768 17.18064229529231, -23.49226042442516 17.083003446192038 C-23.799609662611243 16.969896085434346, -24.106958900797324 16.856788724676655, -24.575118587828737 16.68450163222839 C-24.80804835632084 16.58139053333302, -25.040978124812938 16.478279434437653, -25.630219804116045 16.217439622243546 C-25.981316365630164 16.034272770989713, -26.332412927144283 15.851105919735875, -26.653228436070876 15.683736674221013 C-27.015326064587512 15.464230963892366, -27.37742369310415 15.244725253563718, -27.639940721789863 15.085585888053139 C-28.004850436593003 14.831040667693424, -28.369760151396143 14.576495447333711, -28.586302048841887 14.42544519362322 C-28.768929265897984 14.279804847935331, -28.951556482954082 14.134164502247442, -29.488423615540185 13.706027250644423 C-29.692395285983505 13.520785576342465, -29.896366956426824 13.33554390204051, -30.34259841086275 12.930288301759106 C-30.538630005635678 12.727869585662809, -30.734661600408607 12.525450869566514, -31.145316447355682 12.101416024703703 C-31.406285833940558 11.794866533642342, -31.66725522052543 11.48831704258098, -31.893279184424536 11.222816433457206 C-32.123608830814696 10.914195690756845, -32.35393847720486 10.605574948056486, -32.5834130827452 10.29809988219906 C-32.76354470465276 10.021369338835841, -32.94367632656031 9.744638795472623, -33.21288223409623 9.33106622958946 C-33.43870136706895 8.930101492329605, -33.66452050004167 8.529136755069752, -33.779100014714004 8.325689224335036 C-33.95644340168191 7.957431886755566, -34.13378678864982 7.589174549176095, -34.279739714284624 7.286100176203095 C-34.404239725793225 6.978582740732867, -34.528739737301834 6.67106530526264, -34.71274409689568 6.2165709795835475 C-34.79710145924048 5.9625000447214465, -34.881458821585284 5.708429109859345, -35.07633385465989 5.121496559358584 C-35.173580739791625 4.750652039284668, -35.270827624923356 4.37980751921075, -35.36901491927282 4.005376811213663 C-35.45340768533936 3.5720377938794377, -35.53780045140589 3.138698776545212, -35.5895846014601 2.8727981106008347 C-35.62880868755927 2.568583844487573, -35.66803277365844 2.264369578374311, -35.73713653308556 1.7284144663382852 C-35.762105875778 1.339496799012957, -35.787075218470434 0.9505791316876285, -35.81106439161238 0.5769283962897952 C-35.81106439161238 0.22684664192793552, -35.81106439161238 -0.12323511243392415, -35.81106439161238 -0.5769283962897864 C-35.78808819957357 -0.9348011333426522, -35.76511200753476 -1.292673870395518, -35.73713653308557 -1.7284144663382603 C-35.689801797599436 -2.0955333274679595, -35.6424670621133 -2.4626521885976587, -35.5895846014601 -2.8727981106008262 C-35.51089598065538 -3.2768475475872125, -35.43220735985065 -3.6808969845735993, -35.36901491927283 -4.005376811213655 C-35.30189459554732 -4.261335695711342, -35.234774271821806 -4.517294580209029, -35.07633385465989 -5.121496559358575 C-34.9688360287141 -5.4452628836546815, -34.861338202768316 -5.769029207950788, -34.71274409689569 -6.2165709795835244 C-34.55998904476133 -6.593878911673409, -34.40723399262698 -6.971186843763293, -34.279739714284624 -7.286100176203086 C-34.100830330282456 -7.657609340131165, -33.92192094628029 -8.029118504059243, -33.779100014714004 -8.32568922433503 C-33.662131520196084 -8.53337863069602, -33.545163025678164 -8.741068037057014, -33.21288223409623 -9.331066229589455 C-33.08002677900841 -9.535167885289235, -32.9471713239206 -9.739269540989016, -32.58341308274521 -10.298099882199054 C-32.376907018041045 -10.574799195813121, -32.17040095333688 -10.851498509427188, -31.893279184424543 -11.2228164334572 C-31.703549355198092 -11.445683886799708, -31.51381952597164 -11.668551340142216, -31.145316447355686 -12.101416024703695 C-30.906860596901456 -12.347641267887102, -30.668404746447226 -12.59386651107051, -30.342598410862763 -12.930288301759086 C-30.0199298358887 -13.223327364207869, -29.69726126091464 -13.516366426656651, -29.488423615540206 -13.706027250644407 C-29.197957075921334 -13.937666586351215, -28.907490536302465 -14.16930592205802, -28.586302048841894 -14.425445193623215 C-28.333561164968515 -14.60174629323676, -28.080820281095136 -14.778047392850302, -27.639940721789888 -15.085585888053124 C-27.35332049176486 -15.259336755861732, -27.066700261739825 -15.433087623670339, -26.653228436070883 -15.683736674221008 C-26.31894167405489 -15.858133863843861, -25.9846549120389 -16.032531053466716, -25.630219804116052 -16.217439622243543 C-25.366090610122658 -16.33436178131026, -25.101961416129264 -16.451283940376978, -24.575118587828744 -16.684501632228386 C-24.282983158883454 -16.792010170550554, -23.990847729938164 -16.899518708872726, -23.492260424425186 -17.08300344619203 C-23.083134035303868 -17.204430010935, -22.67400764618255 -17.325856575677964, -22.38609501037115 -17.411307534702523 C-21.93663400696511 -17.513894075722803, -21.48717300355907 -17.61648061674308, -21.26116781662471 -17.668064825839174 C-21.01261016824435 -17.70824969883596, -20.764052519863995 -17.748434571832743, -20.12210141032112 -17.85222024881843 C-19.817176916927377 -17.881635947970768, -19.51225242353363 -17.911051647123106, -18.973576459652833 -17.963017069506055 C-18.578664895822868 -17.975681114709353, -18.183753331992904 -17.98834515991265, -17.820312500000007 -18 C-17.820312500000004 -18, -17.820312500000004 -18, -17.8203125 -18" stroke="none" stroke-width="0" fill="#eee" style=""/><path d="M-17.8203125 -18 C-3.6154075957132807 -18, 10.589497308573439 -18, 17.8203125 -18 M-17.8203125 -18 C-7.398019775222039 -18, 3.0242729495559217 -18, 17.8203125 -18 M17.8203125 -18 C17.8203125 -18, 17.8203125 -18, 17.8203125 -18 M17.8203125 -18 C17.8203125 -18, 17.8203125 -18, 17.8203125 -18 M17.8203125 -18 C18.226556939958904 -17.986972531517353, 18.63280137991781 -17.9739450630347, 18.973576459652833 -17.963017069506055 M17.8203125 -18 C18.2181621462531 -17.98724173621202, 18.616011792506196 -17.974483472424044, 18.973576459652833 -17.963017069506055 M18.973576459652833 -17.963017069506055 C19.258917943731593 -17.93549051971032, 19.544259427810353 -17.907963969914583, 20.122101410321108 -17.85222024881843 M18.973576459652833 -17.963017069506055 C19.411381604976086 -17.920782533674444, 19.849186750299335 -17.87854799784283, 20.122101410321108 -17.85222024881843 M20.122101410321108 -17.85222024881843 C20.412929444237754 -17.805201427699874, 20.703757478154397 -17.758182606581318, 21.2611678166247 -17.668064825839178 M20.122101410321108 -17.85222024881843 C20.526565223369413 -17.786829676188496, 20.931029036417716 -17.721439103558563, 21.2611678166247 -17.668064825839178 M21.2611678166247 -17.668064825839178 C21.48833170262817 -17.616216151241776, 21.71549558863164 -17.564367476644378, 22.38609501037113 -17.41130753470253 M21.2611678166247 -17.668064825839178 C21.493636800728076 -17.615005297219476, 21.726105784831454 -17.56194576859977, 22.38609501037113 -17.41130753470253 M22.38609501037113 -17.41130753470253 C22.74314926073045 -17.30533570741189, 23.100203511089774 -17.199363880121247, 23.492260424425172 -17.083003446192034 M22.38609501037113 -17.41130753470253 C22.66565258696777 -17.32833631339631, 22.945210163564408 -17.245365092090093, 23.492260424425172 -17.083003446192034 M23.492260424425172 -17.083003446192034 C23.867387693359223 -16.944953143585842, 24.242514962293274 -16.80690284097965, 24.575118587828733 -16.68450163222839 M23.492260424425172 -17.083003446192034 C23.758624484131236 -16.984979009977828, 24.024988543837296 -16.886954573763617, 24.575118587828733 -16.68450163222839 M24.575118587828733 -16.68450163222839 C24.89104389973407 -16.5446508682393, 25.206969211639407 -16.404800104250217, 25.630219804116045 -16.217439622243546 M24.575118587828733 -16.68450163222839 C24.831295136263666 -16.571099873443554, 25.087471684698595 -16.457698114658722, 25.630219804116045 -16.217439622243546 M25.630219804116045 -16.217439622243546 C25.962990257065094 -16.043833490089675, 26.295760710014143 -15.870227357935804, 26.653228436070876 -15.683736674221011 M25.630219804116045 -16.217439622243546 C26.025623358470007 -16.011157841205147, 26.42102691282397 -15.80487606016675, 26.653228436070876 -15.683736674221011 M26.653228436070876 -15.683736674221011 C26.990303417779856 -15.479399839436562, 27.327378399488836 -15.275063004652113, 27.639940721789877 -15.085585888053133 M26.653228436070876 -15.683736674221011 C26.85507151118918 -15.561378215922723, 27.056914586307485 -15.439019757624436, 27.639940721789877 -15.085585888053133 M27.639940721789877 -15.085585888053133 C27.94170384714368 -14.875088994921475, 28.243466972497483 -14.664592101789818, 28.586302048841887 -14.425445193623219 M27.639940721789877 -15.085585888053133 C27.887785073575817 -14.91270039725633, 28.135629425361753 -14.739814906459525, 28.586302048841887 -14.425445193623219 M28.586302048841887 -14.425445193623219 C28.793976733098464 -14.259830159384041, 29.001651417355045 -14.094215125144864, 29.488423615540185 -13.706027250644421 M28.586302048841887 -14.425445193623219 C28.8958165785074 -14.178615592742519, 29.205331108172913 -13.931785991861817, 29.488423615540185 -13.706027250644421 M29.488423615540185 -13.706027250644421 C29.698336374015515 -13.51539003738437, 29.90824913249084 -13.324752824124317, 30.342598410862756 -12.930288301759097 M29.488423615540185 -13.706027250644421 C29.77931692307468 -13.44184564535675, 30.070210230609177 -13.177664040069079, 30.342598410862756 -12.930288301759097 M30.342598410862756 -12.930288301759097 C30.520348004892533 -12.746747252410279, 30.69809759892231 -12.563206203061462, 31.145316447355675 -12.101416024703706 M30.342598410862756 -12.930288301759097 C30.588352402735726 -12.676527128372285, 30.834106394608696 -12.422765954985476, 31.145316447355675 -12.101416024703706 M31.145316447355675 -12.101416024703706 C31.321937068998274 -11.893947392524236, 31.49855769064087 -11.686478760344768, 31.893279184424536 -11.222816433457204 M31.145316447355675 -12.101416024703706 C31.419710255820476 -11.77909744383094, 31.694104064285277 -11.456778862958174, 31.893279184424536 -11.222816433457204 M31.893279184424536 -11.222816433457204 C32.078311167436226 -10.97489043298197, 32.26334315044791 -10.726964432506737, 32.5834130827452 -10.298099882199057 M31.893279184424536 -11.222816433457204 C32.09669680885314 -10.950255348449694, 32.30011443328174 -10.677694263442183, 32.5834130827452 -10.298099882199057 M32.5834130827452 -10.298099882199057 C32.81500632858841 -9.94231046508975, 33.04659957443162 -9.586521047980446, 33.21288223409623 -9.331066229589453 M32.5834130827452 -10.298099882199057 C32.80548420252552 -9.956939007771219, 33.02755532230584 -9.615778133343381, 33.21288223409623 -9.331066229589453 M33.21288223409623 -9.331066229589453 C33.41738272595024 -8.967954898382372, 33.621883217804246 -8.604843567175292, 33.779100014714004 -8.325689224335035 M33.21288223409623 -9.331066229589453 C33.334437948292205 -9.115231750183712, 33.455993662488176 -8.899397270777971, 33.779100014714004 -8.325689224335035 M33.779100014714004 -8.325689224335035 C33.93135427521985 -8.009529994675342, 34.0836085357257 -7.693370765015651, 34.279739714284624 -7.286100176203093 M33.779100014714004 -8.325689224335035 C33.87950485152282 -8.11719643238218, 33.97990968833163 -7.9087036404293265, 34.279739714284624 -7.286100176203093 M34.279739714284624 -7.286100176203093 C34.41683166384279 -6.947480410301878, 34.55392361340096 -6.608860644400663, 34.71274409689568 -6.216570979583546 M34.279739714284624 -7.286100176203093 C34.38499041710294 -7.0261289053579485, 34.49024111992125 -6.766157634512804, 34.71274409689568 -6.216570979583546 M34.71274409689568 -6.216570979583546 C34.80500489210184 -5.938696164184051, 34.897265687308 -5.660821348784557, 35.07633385465989 -5.121496559358589 M34.71274409689568 -6.216570979583546 C34.790910496459645 -5.981146238221355, 34.869076896023614 -5.745721496859165, 35.07633385465989 -5.121496559358589 M35.07633385465989 -5.121496559358589 C35.139764754666494 -4.879607046507465, 35.20319565467309 -4.637717533656339, 35.36901491927283 -4.005376811213661 M35.07633385465989 -5.121496559358589 C35.173123427075886 -4.7523959708703725, 35.26991299949188 -4.383295382382155, 35.36901491927283 -4.005376811213661 M35.36901491927283 -4.005376811213661 C35.44188534277973 -3.6312025876751783, 35.51475576628663 -3.257028364136696, 35.5895846014601 -2.872798110600833 M35.36901491927283 -4.005376811213661 C35.42949171669368 -3.6948412382332307, 35.48996851411454 -3.384305665252801, 35.5895846014601 -2.872798110600833 M35.5895846014601 -2.872798110600833 C35.629337820181796 -2.5644799963563076, 35.669091038903495 -2.256161882111783, 35.73713653308557 -1.728414466338275 M35.5895846014601 -2.872798110600833 C35.62185572748835 -2.6225096297803385, 35.65412685351661 -2.372221148959844, 35.73713653308557 -1.728414466338275 M35.73713653308557 -1.728414466338275 C35.75280681342587 -1.484337201215443, 35.76847709376617 -1.240259936092611, 35.81106439161238 -0.576928396289793 M35.73713653308557 -1.728414466338275 C35.752783294895224 -1.4847035213140454, 35.76843005670489 -1.2409925762898155, 35.81106439161238 -0.576928396289793 M35.81106439161238 -0.576928396289793 C35.81106439161238 -0.14769429765257924, 35.81106439161238 0.2815398009846345, 35.81106439161238 0.5769283962897886 M35.81106439161238 -0.576928396289793 C35.81106439161238 -0.1847732309955737, 35.81106439161238 0.2073819342986456, 35.81106439161238 0.5769283962897886 M35.81106439161238 0.5769283962897886 C35.782535475511146 1.021289292693194, 35.7540065594099 1.4656501890965994, 35.73713653308557 1.7284144663382628 M35.81106439161238 0.5769283962897886 C35.79613099270642 0.8095281382727743, 35.781197593800464 1.0421278802557599, 35.73713653308557 1.7284144663382628 M35.73713653308557 1.7284144663382628 C35.696910017162544 2.040403380213184, 35.65668350123952 2.3523922940881055, 35.5895846014601 2.8727981106008205 M35.73713653308557 1.7284144663382628 C35.70760011738601 1.9574930738790592, 35.67806370168645 2.1865716814198555, 35.5895846014601 2.8727981106008205 M35.5895846014601 2.8727981106008205 C35.521845418981684 3.2206244959462227, 35.454106236503264 3.568450881291625, 35.36901491927283 4.005376811213657 M35.5895846014601 2.8727981106008205 C35.534468900228894 3.155805587715512, 35.47935319899768 3.438813064830203, 35.36901491927283 4.005376811213657 M35.36901491927283 4.005376811213657 C35.3039305908837 4.253571563087751, 35.238846262494576 4.501766314961844, 35.07633385465989 5.121496559358577 M35.36901491927283 4.005376811213657 C35.28981419588297 4.307403495309829, 35.210613472493115 4.609430179406001, 35.07633385465989 5.121496559358577 M35.07633385465989 5.121496559358577 C35.000630895102454 5.349501811916972, 34.92492793554502 5.577507064475368, 34.71274409689569 6.216570979583534 M35.07633385465989 5.121496559358577 C35.00325610019697 5.341595112441923, 34.93017834573405 5.561693665525269, 34.71274409689569 6.216570979583534 M34.71274409689569 6.216570979583534 C34.542454569264144 6.637189405157499, 34.37216504163259 7.057807830731464, 34.279739714284624 7.286100176203088 M34.71274409689569 6.216570979583534 C34.57337358598899 6.560818836068602, 34.43400307508229 6.905066692553671, 34.279739714284624 7.286100176203088 M34.279739714284624 7.286100176203088 C34.083532497457085 7.693528660107385, 33.88732528062954 8.100957144011684, 33.779100014714004 8.325689224335031 M34.279739714284624 7.286100176203088 C34.146927540736506 7.561887496301038, 34.01411536718838 7.837674816398988, 33.779100014714004 8.325689224335031 M33.779100014714004 8.325689224335031 C33.62343475901204 8.60208864865027, 33.46776950331007 8.878488072965506, 33.21288223409623 9.33106622958945 M33.779100014714004 8.325689224335031 C33.590028290890245 8.661405215572223, 33.40095656706649 8.997121206809414, 33.21288223409623 9.33106622958945 M33.21288223409623 9.33106622958945 C32.99893459137532 9.659747261795022, 32.7849869486544 9.988428294000595, 32.58341308274521 10.298099882199049 M33.21288223409623 9.33106622958945 C33.07660185488632 9.540429488461669, 32.94032147567641 9.749792747333887, 32.58341308274521 10.298099882199049 M32.58341308274521 10.298099882199049 C32.40059637330079 10.543057617909716, 32.21777966385637 10.78801535362038, 31.89327918442454 11.2228164334572 M32.58341308274521 10.298099882199049 C32.40834623128228 10.53267351424111, 32.23327937981934 10.767247146283172, 31.89327918442454 11.2228164334572 M31.89327918442454 11.2228164334572 C31.59758970115178 11.570150125717616, 31.301900217879023 11.917483817978033, 31.145316447355682 12.101416024703695 M31.89327918442454 11.2228164334572 C31.675527042985166 11.478600489239742, 31.457774901545793 11.734384545022282, 31.145316447355682 12.101416024703695 M31.145316447355682 12.101416024703695 C30.831128007274167 12.42584138443852, 30.51693956719265 12.750266744173347, 30.34259841086276 12.930288301759093 M31.145316447355682 12.101416024703695 C30.898127720106014 12.356658680160914, 30.650938992856346 12.611901335618134, 30.34259841086276 12.930288301759093 M30.34259841086276 12.930288301759093 C30.006122105100527 13.235867179977156, 29.66964579933829 13.541446058195218, 29.488423615540203 13.706027250644409 M30.34259841086276 12.930288301759093 C30.122146496600735 13.130496897453844, 29.901694582338706 13.330705493148594, 29.488423615540203 13.706027250644409 M29.488423615540203 13.706027250644409 C29.298028518641473 13.857862273794833, 29.107633421742744 14.009697296945257, 28.586302048841894 14.425445193623217 M29.488423615540203 13.706027250644409 C29.30498026558308 13.852318440602659, 29.121536915625953 13.998609630560908, 28.586302048841894 14.425445193623217 M28.586302048841894 14.425445193623217 C28.286107785013424 14.634847716869722, 27.985913521184955 14.844250240116224, 27.639940721789884 15.085585888053126 M28.586302048841894 14.425445193623217 C28.354240927919445 14.587320985814701, 28.122179806996993 14.749196778006187, 27.639940721789884 15.085585888053126 M27.639940721789884 15.085585888053126 C27.401054660867334 15.230400022034695, 27.162168599944785 15.375214156016265, 26.653228436070883 15.68373667422101 M27.639940721789884 15.085585888053126 C27.294700639079412 15.294872455048141, 26.949460556368937 15.504159022043156, 26.653228436070883 15.68373667422101 M26.653228436070883 15.68373667422101 C26.4249158573435 15.802847200301898, 26.19660327861612 15.921957726382788, 25.63021980411605 16.217439622243543 M26.653228436070883 15.68373667422101 C26.27277839999297 15.882217214352977, 25.892328363915055 16.080697754484945, 25.63021980411605 16.217439622243543 M25.63021980411605 16.217439622243543 C25.34335637371326 16.344425553381537, 25.056492943310474 16.471411484519532, 24.57511858782874 16.68450163222839 M25.63021980411605 16.217439622243543 C25.414825939081698 16.31278809686226, 25.199432074047348 16.408136571480977, 24.57511858782874 16.68450163222839 M24.57511858782874 16.68450163222839 C24.19187666312812 16.825538203167735, 23.8086347384275 16.96657477410708, 23.492260424425183 17.08300344619203 M24.57511858782874 16.68450163222839 C24.266007081379108 16.798257523997297, 23.956895574929476 16.912013415766204, 23.492260424425183 17.08300344619203 M23.492260424425183 17.08300344619203 C23.128516480866438 17.190960739309347, 22.764772537307692 17.29891803242666, 22.38609501037115 17.411307534702527 M23.492260424425183 17.08300344619203 C23.056456100113948 17.212347882064027, 22.620651775802717 17.341692317936023, 22.38609501037115 17.411307534702527 M22.38609501037115 17.411307534702527 C21.967789171510276 17.506783112721816, 21.5494833326494 17.60225869074111, 21.261167816624706 17.668064825839174 M22.38609501037115 17.411307534702527 C22.050696823516226 17.487859982174452, 21.715298636661302 17.564412429646374, 21.261167816624706 17.668064825839174 M21.261167816624706 17.668064825839174 C21.01410481090874 17.708008056601106, 20.767041805192775 17.74795128736304, 20.122101410321118 17.85222024881843 M21.261167816624706 17.668064825839174 C20.825406587341554 17.738515322010873, 20.3896453580584 17.808965818182568, 20.122101410321118 17.85222024881843 M20.122101410321118 17.85222024881843 C19.85565695339579 17.877923825131674, 19.589212496470466 17.90362740144492, 18.973576459652833 17.963017069506055 M20.122101410321118 17.85222024881843 C19.689627730258024 17.89394046460842, 19.25715405019493 17.93566068039841, 18.973576459652833 17.963017069506055 M18.973576459652833 17.963017069506055 C18.70295578630981 17.97169534776898, 18.432335112966793 17.980373626031906, 17.820312500000004 18 M18.973576459652833 17.963017069506055 C18.590889020016444 17.97528911083928, 18.20820158038006 17.987561152172503, 17.820312500000004 18 M17.820312500000004 18 C17.820312500000004 18, 17.820312500000004 18, 17.8203125 18 M17.820312500000004 18 C17.820312500000004 18, 17.8203125 18, 17.8203125 18 M17.8203125 18 C6.045363945257639 18, -5.729584609484721 18, -17.820312499999996 18 M17.8203125 18 C5.546869551259908 18, -6.726573397480184 18, -17.820312499999996 18 M-17.820312499999996 18 C-18.149572398803123 17.9894412759109, -18.478832297606253 17.978882551821798, -18.973576459652826 17.963017069506055 M-17.820312499999996 18 C-18.195373173895785 17.98797253419948, -18.570433847791573 17.975945068398953, -18.973576459652826 17.963017069506055 M-18.973576459652826 17.963017069506055 C-19.372603104877932 17.924523448940942, -19.771629750103042 17.88602982837583, -20.12210141032111 17.85222024881843 M-18.973576459652826 17.963017069506055 C-19.352758655265756 17.926437819128587, -19.731940850878686 17.889858568751116, -20.12210141032111 17.85222024881843 M-20.12210141032111 17.85222024881843 C-20.508738474763614 17.78971176665458, -20.895375539206118 17.727203284490734, -21.2611678166247 17.668064825839174 M-20.12210141032111 17.85222024881843 C-20.570735919577036 17.779688500574512, -21.01937042883296 17.707156752330594, -21.2611678166247 17.668064825839174 M-21.2611678166247 17.668064825839174 C-21.57756898677632 17.59584832346266, -21.893970156927942 17.523631821086145, -22.38609501037113 17.41130753470253 M-21.2611678166247 17.668064825839174 C-21.638208520662438 17.582007745563093, -22.01524922470018 17.495950665287012, -22.38609501037113 17.41130753470253 M-22.38609501037113 17.41130753470253 C-22.628280552254576 17.339428137129374, -22.870466094138028 17.267548739556215, -23.49226042442516 17.083003446192038 M-22.38609501037113 17.41130753470253 C-22.70611843784761 17.31632626126476, -23.026141865324092 17.221344987826992, -23.49226042442516 17.083003446192038 M-23.49226042442516 17.083003446192038 C-23.904917824035415 16.931141708572277, -24.317575223645672 16.77927997095252, -24.575118587828737 16.68450163222839 M-23.49226042442516 17.083003446192038 C-23.762045299574886 16.983720118272892, -24.031830174724607 16.884436790353746, -24.575118587828737 16.68450163222839 M-24.575118587828737 16.68450163222839 C-24.845741253665384 16.56470500579866, -25.11636391950203 16.444908379368936, -25.630219804116045 16.217439622243546 M-24.575118587828737 16.68450163222839 C-24.876423741532957 16.551122778583352, -25.177728895237173 16.417743924938318, -25.630219804116045 16.217439622243546 M-25.630219804116045 16.217439622243546 C-25.89324550475614 16.08021928269561, -26.156271205396237 15.942998943147675, -26.653228436070876 15.683736674221013 M-25.630219804116045 16.217439622243546 C-25.898785198071597 16.07732922824128, -26.167350592027145 15.937218834239015, -26.653228436070876 15.683736674221013 M-26.653228436070876 15.683736674221013 C-26.922042158675662 15.52078021585321, -27.19085588128045 15.357823757485406, -27.639940721789863 15.085585888053139 M-26.653228436070876 15.683736674221013 C-27.022554588250625 15.459848990368267, -27.391880740430373 15.235961306515522, -27.639940721789863 15.085585888053139 M-27.639940721789863 15.085585888053139 C-27.849319796565815 14.939532109363498, -28.058698871341768 14.793478330673857, -28.586302048841887 14.42544519362322 M-27.639940721789863 15.085585888053139 C-27.904135180730776 14.901295270289758, -28.16832963967169 14.717004652526375, -28.586302048841887 14.42544519362322 M-28.586302048841887 14.42544519362322 C-28.91344657655997 14.164556138449546, -29.240591104278053 13.903667083275872, -29.488423615540185 13.706027250644423 M-28.586302048841887 14.42544519362322 C-28.811368050103866 14.24596104687462, -29.036434051365845 14.066476900126016, -29.488423615540185 13.706027250644423 M-29.488423615540185 13.706027250644423 C-29.789819146504506 13.43230780391139, -30.091214677468823 13.15858835717836, -30.34259841086275 12.930288301759106 M-29.488423615540185 13.706027250644423 C-29.810365960959075 13.413647730662671, -30.13230830637796 13.12126821068092, -30.34259841086275 12.930288301759106 M-30.34259841086275 12.930288301759106 C-30.545759479542244 12.720507818513584, -30.748920548221744 12.510727335268061, -31.145316447355682 12.101416024703703 M-30.34259841086275 12.930288301759106 C-30.64918393834762 12.613713573491053, -30.955769465832486 12.297138845223001, -31.145316447355682 12.101416024703703 M-31.145316447355682 12.101416024703703 C-31.4387982245541 11.756675629620243, -31.73228000175251 11.411935234536783, -31.893279184424536 11.222816433457206 M-31.145316447355682 12.101416024703703 C-31.432647182100737 11.763900994090191, -31.719977916845792 11.42638596347668, -31.893279184424536 11.222816433457206 M-31.893279184424536 11.222816433457206 C-32.10037104354405 10.945332208693538, -32.30746290266357 10.66784798392987, -32.5834130827452 10.29809988219906 M-31.893279184424536 11.222816433457206 C-32.037366350183206 11.029752757951426, -32.18145351594188 10.836689082445647, -32.5834130827452 10.29809988219906 M-32.5834130827452 10.29809988219906 C-32.786609998899735 9.985934852973749, -32.98980691505427 9.67376982374844, -33.21288223409623 9.33106622958946 M-32.5834130827452 10.29809988219906 C-32.72506366837037 10.08048654417067, -32.86671425399553 9.862873206142282, -33.21288223409623 9.33106622958946 M-33.21288223409623 9.33106622958946 C-33.366778988868894 9.05780695775433, -33.520675743641554 8.784547685919202, -33.779100014714004 8.325689224335036 M-33.21288223409623 9.33106622958946 C-33.34810440827072 9.090965566612724, -33.483326582445216 8.850864903635985, -33.779100014714004 8.325689224335036 M-33.779100014714004 8.325689224335036 C-33.88320462296945 8.109513777810811, -33.987309231224906 7.893338331286587, -34.279739714284624 7.286100176203095 M-33.779100014714004 8.325689224335036 C-33.948414965951024 7.9741031053317695, -34.117729917188036 7.6225169863285025, -34.279739714284624 7.286100176203095 M-34.279739714284624 7.286100176203095 C-34.442151924013096 6.884938878147307, -34.60456413374157 6.483777580091518, -34.71274409689568 6.2165709795835475 M-34.279739714284624 7.286100176203095 C-34.37220252223557 7.05771525291749, -34.464665330186506 6.829330329631885, -34.71274409689568 6.2165709795835475 M-34.71274409689568 6.2165709795835475 C-34.82936957945983 5.865313607847547, -34.94599506202397 5.514056236111546, -35.07633385465989 5.121496559358584 M-34.71274409689568 6.2165709795835475 C-34.79697813240519 5.9628714854998925, -34.88121216791469 5.709171991416238, -35.07633385465989 5.121496559358584 M-35.07633385465989 5.121496559358584 C-35.189253523491736 4.690884922408721, -35.30217319232357 4.260273285458856, -35.36901491927282 4.005376811213663 M-35.07633385465989 5.121496559358584 C-35.19104211690854 4.684064240529133, -35.305750379157196 4.246631921699682, -35.36901491927282 4.005376811213663 M-35.36901491927282 4.005376811213663 C-35.41656521071168 3.7612161144327616, -35.46411550215055 3.5170554176518594, -35.5895846014601 2.8727981106008347 M-35.36901491927282 4.005376811213663 C-35.4199731516273 3.7437170580996275, -35.470931383981785 3.4820573049855916, -35.5895846014601 2.8727981106008347 M-35.5895846014601 2.8727981106008347 C-35.644981532882575 2.4431504498495666, -35.70037846430505 2.013502789098299, -35.73713653308556 1.7284144663382852 M-35.5895846014601 2.8727981106008347 C-35.641068800368274 2.473496830574034, -35.69255299927645 2.074195550547233, -35.73713653308556 1.7284144663382852 M-35.73713653308556 1.7284144663382852 C-35.75912937784544 1.3858581568984252, -35.78112222260532 1.0433018474585654, -35.81106439161238 0.5769283962897952 M-35.73713653308556 1.7284144663382852 C-35.765354416891995 1.2888981489872524, -35.79357230069843 0.8493818316362198, -35.81106439161238 0.5769283962897952 M-35.81106439161238 0.5769283962897952 C-35.81106439161238 0.27025518773281526, -35.81106439161238 -0.036418020824164676, -35.81106439161238 -0.5769283962897864 M-35.81106439161238 0.5769283962897952 C-35.81106439161238 0.13653269028117426, -35.81106439161238 -0.3038630157274467, -35.81106439161238 -0.5769283962897864 M-35.81106439161238 -0.5769283962897864 C-35.782493212633966 -1.0219475711191914, -35.75392203365555 -1.4669667459485964, -35.73713653308557 -1.7284144663382603 M-35.81106439161238 -0.5769283962897864 C-35.78353463949299 -1.0057265071652233, -35.75600488737361 -1.4345246180406603, -35.73713653308557 -1.7284144663382603 M-35.73713653308557 -1.7284144663382603 C-35.69599968780889 -2.0474637149152644, -35.65486284253221 -2.3665129634922684, -35.5895846014601 -2.8727981106008262 M-35.73713653308557 -1.7284144663382603 C-35.686772206729984 -2.1190302359910027, -35.636407880374406 -2.509646005643745, -35.5895846014601 -2.8727981106008262 M-35.5895846014601 -2.8727981106008262 C-35.539317693590874 -3.1309080585945672, -35.48905078572165 -3.3890180065883087, -35.36901491927283 -4.005376811213655 M-35.5895846014601 -2.8727981106008262 C-35.54347256065789 -3.1095736943369086, -35.49736051985568 -3.346349278072991, -35.36901491927283 -4.005376811213655 M-35.36901491927283 -4.005376811213655 C-35.30025418704199 -4.267591284299681, -35.231493454811165 -4.529805757385707, -35.07633385465989 -5.121496559358575 M-35.36901491927283 -4.005376811213655 C-35.29622556179938 -4.282954178366985, -35.223436204325935 -4.560531545520315, -35.07633385465989 -5.121496559358575 M-35.07633385465989 -5.121496559358575 C-34.988659369026784 -5.3855581410532505, -34.90098488339367 -5.649619722747926, -34.71274409689569 -6.2165709795835244 M-35.07633385465989 -5.121496559358575 C-34.94046442799599 -5.530713624616952, -34.80459500133208 -5.939930689875331, -34.71274409689569 -6.2165709795835244 M-34.71274409689569 -6.2165709795835244 C-34.58461491217113 -6.53305254306245, -34.45648572744657 -6.849534106541375, -34.279739714284624 -7.286100176203086 M-34.71274409689569 -6.2165709795835244 C-34.60826015587445 -6.474648333965577, -34.5037762148532 -6.73272568834763, -34.279739714284624 -7.286100176203086 M-34.279739714284624 -7.286100176203086 C-34.085699387706406 -7.689029066140643, -33.89165906112819 -8.091957956078199, -33.779100014714004 -8.32568922433503 M-34.279739714284624 -7.286100176203086 C-34.12131351376081 -7.615075571368447, -33.962887313237 -7.9440509665338075, -33.779100014714004 -8.32568922433503 M-33.779100014714004 -8.32568922433503 C-33.6378934858018 -8.576415715130201, -33.4966869568896 -8.827142205925371, -33.21288223409623 -9.331066229589455 M-33.779100014714004 -8.32568922433503 C-33.58264614897225 -8.67451295607055, -33.3861922832305 -9.023336687806072, -33.21288223409623 -9.331066229589455 M-33.21288223409623 -9.331066229589455 C-32.972331411519825 -9.700616890007094, -32.73178058894343 -10.070167550424731, -32.58341308274521 -10.298099882199054 M-33.21288223409623 -9.331066229589455 C-32.96426470332419 -9.713009520010047, -32.715647172552146 -10.094952810430637, -32.58341308274521 -10.298099882199054 M-32.58341308274521 -10.298099882199054 C-32.41605681586246 -10.52234203335237, -32.24870054897972 -10.746584184505686, -31.893279184424543 -11.2228164334572 M-32.58341308274521 -10.298099882199054 C-32.393921642545806 -10.55200114904092, -32.2044302023464 -10.805902415882787, -31.893279184424543 -11.2228164334572 M-31.893279184424543 -11.2228164334572 C-31.606349409795868 -11.559860473502734, -31.319419635167193 -11.896904513548266, -31.145316447355686 -12.101416024703695 M-31.893279184424543 -11.2228164334572 C-31.630706307788532 -11.53124947517379, -31.368133431152526 -11.839682516890383, -31.145316447355686 -12.101416024703695 M-31.145316447355686 -12.101416024703695 C-30.94029042124558 -12.313122229612171, -30.735264395135477 -12.524828434520646, -30.342598410862763 -12.930288301759086 M-31.145316447355686 -12.101416024703695 C-30.956781858792475 -12.296093466372803, -30.768247270229264 -12.490770908041908, -30.342598410862763 -12.930288301759086 M-30.342598410862763 -12.930288301759086 C-30.07897343187001 -13.169705532545587, -29.81534845287726 -13.409122763332087, -29.488423615540206 -13.706027250644407 M-30.342598410862763 -12.930288301759086 C-30.05785665685987 -13.188883228890546, -29.773114902856978 -13.447478156022008, -29.488423615540206 -13.706027250644407 M-29.488423615540206 -13.706027250644407 C-29.153904752162042 -13.972797142267682, -28.81938588878388 -14.239567033890957, -28.586302048841894 -14.425445193623215 M-29.488423615540206 -13.706027250644407 C-29.290979959656422 -13.86348331201538, -29.093536303772638 -14.020939373386353, -28.586302048841894 -14.425445193623215 M-28.586302048841894 -14.425445193623215 C-28.271225043437767 -14.645229273129399, -27.95614803803364 -14.865013352635584, -27.639940721789888 -15.085585888053124 M-28.586302048841894 -14.425445193623215 C-28.387993291723568 -14.563776797941223, -28.189684534605245 -14.702108402259231, -27.639940721789888 -15.085585888053124 M-27.639940721789888 -15.085585888053124 C-27.433230626949747 -15.210894762308572, -27.226520532109607 -15.336203636564019, -26.653228436070883 -15.683736674221008 M-27.639940721789888 -15.085585888053124 C-27.427221199016955 -15.214537712836748, -27.214501676244023 -15.343489537620373, -26.653228436070883 -15.683736674221008 M-26.653228436070883 -15.683736674221008 C-26.431240188796018 -15.799547800634047, -26.209251941521153 -15.915358927047086, -25.630219804116052 -16.217439622243543 M-26.653228436070883 -15.683736674221008 C-26.365645777124385 -15.833768363187911, -26.078063118177887 -15.983800052154814, -25.630219804116052 -16.217439622243543 M-25.630219804116052 -16.217439622243543 C-25.305909207747142 -16.361002302924028, -24.98159861137823 -16.504564983604514, -24.575118587828744 -16.684501632228386 M-25.630219804116052 -16.217439622243543 C-25.36386282435783 -16.335347955983433, -25.09750584459961 -16.45325628972332, -24.575118587828744 -16.684501632228386 M-24.575118587828744 -16.684501632228386 C-24.30611624192265 -16.783496982152258, -24.037113896016553 -16.882492332076126, -23.492260424425186 -17.08300344619203 M-24.575118587828744 -16.684501632228386 C-24.174145324013452 -16.83206349986863, -23.773172060198164 -16.97962536750888, -23.492260424425186 -17.08300344619203 M-23.492260424425186 -17.08300344619203 C-23.051914880577495 -17.213695692185098, -22.611569336729804 -17.344387938178166, -22.38609501037115 -17.411307534702523 M-23.492260424425186 -17.08300344619203 C-23.09101980788064 -17.202089555027477, -22.689779191336097 -17.32117566386292, -22.38609501037115 -17.411307534702523 M-22.38609501037115 -17.411307534702523 C-22.018962508989983 -17.49510313237931, -21.651830007608815 -17.578898730056093, -21.26116781662471 -17.668064825839174 M-22.38609501037115 -17.411307534702523 C-21.96617442884179 -17.507151667198716, -21.546253847312432 -17.60299579969491, -21.26116781662471 -17.668064825839174 M-21.26116781662471 -17.668064825839174 C-20.835653089720736 -17.736858746969016, -20.410138362816767 -17.805652668098862, -20.12210141032112 -17.85222024881843 M-21.26116781662471 -17.668064825839174 C-20.891567741433494 -17.7278188997047, -20.521967666242276 -17.787572973570224, -20.12210141032112 -17.85222024881843 M-20.12210141032112 -17.85222024881843 C-19.678659606201215 -17.894998546345082, -19.235217802081305 -17.93777684387173, -18.973576459652833 -17.963017069506055 M-20.12210141032112 -17.85222024881843 C-19.670102515569102 -17.895824038586955, -19.218103620817082 -17.939427828355477, -18.973576459652833 -17.963017069506055 M-18.973576459652833 -17.963017069506055 C-18.5918031817374 -17.975259795452196, -18.21002990382197 -17.987502521398337, -17.820312500000007 -18 M-18.973576459652833 -17.963017069506055 C-18.641579110324262 -17.973663578305384, -18.30958176099569 -17.984310087104713, -17.820312500000007 -18 M-17.820312500000007 -18 C-17.820312500000004 -18, -17.820312500000004 -18, -17.8203125 -18 M-17.820312500000007 -18 C-17.820312500000004 -18, -17.820312500000004 -18, -17.8203125 -18" stroke="#999" stroke-width="1.3" fill="none" stroke-dasharray="0 0" style=""/></g><g class="label" style="" transform="translate(-23.8203125, -10.5)"><rect/><foreignObject width="47.640625" height="21"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel"><p>fulfilled</p></span></div></foreignObject></g></g></g></g></g><defs><filter id="dg3-drop-shadow" height="130%" width="130%"><feDropShadow dx="4" dy="4" stdDeviation="0" flood-opacity="0.06" flood-color="#000000"/></filter></defs><defs><filter id="dg3-drop-shadow-small" height="150%" width="150%"><feDropShadow dx="2" dy="2" stdDeviation="0" flood-opacity="0.06" flood-color="#000000"/></filter></defs><linearGradient id="dg3-gradient" gradientUnits="objectBoundingBox" x1="0%" y1="0%" x2="100%" y2="0%"><stop offset="0%" stop-color="hsl(0, 0%, 83.3333333333%)" stop-opacity="1"/><stop offset="100%" stop-color="hsl(0, 0%, 88.9215686275%)" stop-opacity="1"/></linearGradient></svg></div>
+ <figcaption>Where the gate acts: local make targets → the commit → the PR → CI → the merge. The merge edge matters: after any merge the digest is recomputed over the <em>merged</em> tree, so an ack that only covered one side fails until the merged content is reviewed as a whole.</figcaption>
+</figure>
+
+<h3>Where it acts, surface by surface</h3>
+
+<table>
+ <thead><tr><th>Surface</th><th>What happens there</th></tr></thead>
+ <tbody>
+ <tr><td><strong>Make targets</strong></td><td><code>make drift-check</code> (<code>dc</code>) is the pass/fail gate; <code>make drift-plan</code> (<code>dp</code>) prints one packet per open contract — what changed, what to review, the previous reviewer's rationale; <code>make drift-ack CONTRACT=… RATIONALE="…"</code> (<code>da</code>) records the review, running the contract's <code>verify_commands</code> first.</td></tr>
+ <tr><td><strong>Quality checks</strong></td><td><code>drift-check</code> runs inside the <code>make check</code> aggregate — a full local check cannot pass with an open contract. It also hard-fails on manifest rot (dead globs, orphan acks, shadowed triggers), so a rename that kills a trigger pattern is caught on the same PR.</td></tr>
+ <tr><td><strong>Commits</strong></td><td><code>drift ack</code> writes the ack file <em>and stages it</em>, into the same index the check reads — so the code change, the doc fix, and the ack land in one commit, reviewed as one diff. If staging fails, the ack is rolled back rather than left as a false green.</td></tr>
+ <tr><td><strong>Branches / merges</strong></td><td>Two branches re-acking the same contract may or may not conflict textually — irrelevant either way, because after the merge the digest is recomputed over the merged tree, and a spliced or stale ack simply fails until re-acked.</td></tr>
+ <tr><td><strong>PRs / CI</strong></td><td>A <code>lint-drift</code> CI job runs the check on every PR — currently <em>advisory</em> (visible, not merge-blocking; promotion to a required check is planned once the open-PR backlog lands). The committed ack appears in the PR diff, so the reviewer sees the rationale right next to the change it covers.</td></tr>
+ </tbody>
+</table>
+
+<h3>The pieces on disk</h3>
+
+<ul>
+ <li><code>drift.toml</code> (repo root, human-authored) — the manifest. Each contract: <code>triggers</code> (globs), optional <code>exclude</code>, <code>review</code> targets, optional <code>verify_commands</code>. Live today: <strong>config-docs</strong> (config model + shipped defaults → configuration docs, verified by a boot test), <strong>cli-docs</strong> (CLI source → CLI docs + the agent-CLI contract doc), <strong>keyword-only-convention</strong> (the guard's source → its specification page).</li>
+ <li><code>.drift/acks/&lt;contract-id&gt;.toml</code> (tool-written, committed) — the last fulfilled review: digest, reviewer, timestamp, rationale, per-file trigger snapshot.</li>
+ <li><code>pipelex-dev drift plan|check|ack</code> — the engine the make targets wrap.</li>
+</ul>
+
+<details>
+<summary>Why there is deliberately no bypass</summary>
+<div class="body">
+<p><code>drift ack</code> has no <code>--skip-verify</code> and <code>drift check</code> has no override label. Every escape hatch is a rubber-stamp invitation — and the legitimate escape already exists: acking with a rationale that says <em>“reviewed, no doc change needed”</em> is cheap, honest, and auditable. If a contract opens so often that acking feels like ceremony, the contract is mis-scoped: narrow its triggers, or mechanize it into a derived check. The manifest is kept deliberately small — contracts are added where staleness has actually hurt, and the campaign's density data is exactly the evidence for where that is.</p>
+<p>One hardening detail worth knowing: for a contract with <code>verify_commands</code>, an <em>unstaged</em> edit or untracked file matching the triggers is a hard error, not a warning — the verify commands run on the working tree, so they would otherwise certify content the digest doesn't cover. The check runs both before the verify commands (fail fast) and after them (catching files the verify commands themselves regenerate).</p>
+</div>
+</details>
+
+<h2>How the two halves reinforce each other</h2>
+
+<p>The campaign wasn't just cleanup — it was an evidence-gathering run for the contract system. Every finding logged an observation about whether a contract or a derived check would have caught it. The resulting shortlist feeds the system's pilot verdict:</p>
+
+<ul>
+ <li><strong>Best derived-check candidates:</strong> diff documented closed enumerations against the code's registries; validate doc bundle examples against the real parser (with placeholder scaffolding injected); import-and-check every Python symbol a doc references.</li>
+ <li><strong>Best <em>contract</em> candidate:</strong> make-targets-and-tooling — every documented target <em>exists</em>, so no script scores a single point; what drifts is behavior under a stable name. Un-mechanizable, judgment-only: exactly what a review contract is for.</li>
+ <li><strong>Known limit of any snippet checker</strong> (proven the hard way): it verifies the snippet, never the sentence beside it. A green check must not be read as a clean page.</li>
+</ul>
+
+<details>
+<summary>Open follow-ups after this recap</summary>
+<div class="body">
+<ul>
+ <li><strong>Pilot verdict per contract</strong> (keep / narrow / mechanize / drop), fed by the campaign's shortlist and the per-ack dogfood log.</li>
+ <li><strong>Promote the CI job from advisory to required</strong>, once the open-PR backlog has landed (a required check would instantly fail every PR predating its acks).</li>
+ <li><strong>Deferred code-side findings</strong> — the hunt surfaced defects whose fix belongs in code, not docs (e.g. an error message whose remediation names config fields that don't exist; a dependency floor that admits a parser too old for the documented syntax). They're recorded with evidence and candidate fixes, to be picked up as ordinary follow-ups.</li>
+ <li><strong>The cookbook sweep</strong>, if ever revived — the reusable review harness (rules, carve-outs, verification pipeline) is archived and ready.</li>
+</ul>
+</div>
+</details>
+
+<footer>
+<p>Where things live in the repo: the system doc is <code>docs/contribute/drift-contracts.md</code>; the manifest is <code>drift.toml</code> with acks under <code>.drift/acks/</code>; the campaign record is <code>wip/drift-hunt/findings/SUMMARY.md</code> with the code-side handoff in <code>wip/drift-hunt/deferred-code-findings.md</code>; the full per-finding audit trail is preserved under the pushed git tag <code>drift-hunt-archive</code>.</p>
+<p>Self-contained document — diagrams pre-rendered from mermaid sources; no external resources.</p>
+</footer>
+
+</main>
+</body>
+</html>
diff --git a/wip/drift-hunt/findings/SUMMARY.md b/wip/drift-hunt/findings/SUMMARY.md
new file mode 100644
index 000000000..84d945779
--- /dev/null
+++ b/wip/drift-hunt/findings/SUMMARY.md
@@ -0,0 +1,125 @@
+# Drift Hunt — findings SUMMARY (campaign record)
+
+**CAMPAIGN COMPLETE (2026-07-12).** This file and `inventory.md` are the surviving record; the per-section findings files (with every finding's `file:line` evidence and fix-time resolution records) and the prescreen artifacts were pruned at close-out and live in git history under the pushed **`drift-hunt-archive` tag** — `git show drift-hunt-archive:wip/drift-hunt/findings/<file>` (the tag pins commit `9a6201a5e` and survives squash-merges and branch deletion). The close-out section at the bottom is the handoff: coverage disclosure, Stage-2 resolution totals, the drift-contract shortlist for the Phase 3 verdict, and the deferred code findings.
+
+Per-section defect density is the campaign's headline metric; the rejected-findings column tracks how much the adversarial-verify pass (D6) filtered. Denominators come from `wip/drift-hunt/inventory.md`.
+
+## Per-section scoreboard
+
+Counts are **drift** findings (after the authoring-convention carve-out, below). Convention reclassifications are tracked in the last column.
+
+| Section | Pages | Claims checked | Drift | Convention | Rejected (D6) | Density | 🔴 breaks | 🟠 wrong | 🟡 misleading | ⚪ stale | Findings file |
+|---|---|---|---|---|---|---|---|---|---|---|---|
+| building-methods | 35 | 452 | 44 | 3 | 0 | 1.26 | 25 | 11 | 6 | 2 | `building-methods.md` (pruned — git history) |
+| under-the-hood + advanced | 26 | 435 | 42 | 0 | 0 | 1.62 | 10 | 23 | 4 | 5 | `under-the-hood-advanced.md` (pruned — git history) |
+| features + reliability | 25 | 308 | 10 | 0 | 0 | **0.40** | 2 | 7 | 1 | 0 | `features-reliability.md` (pruned — git history) |
+| the tail (get-started, contribute, tools non-CLI, setup, viewpoint, root, D9 adds) | 18 | 244 | 13 | 0 | 0 | 0.72 | 2 | 9 | 2 | 0 | `the-tail.md` (pruned — git history) |
+| cookbook | 21 | — | — | — | — | — | — | — | — | — | *(NEVER SWEPT — deferred, D12)* |
+| **Total (STAGE 1 COMPLETE)** | **104** | **1439** | **109** | **3** | **0** | **1.05** | 39 | 50 | 13 | 7 | |
+
+> **Part 2 density caveat:** 42/26 = 1.62, but 5 of the 26 pages are 8-line "Under Construction" stubs (`advanced/*-injection.md`, nothing checkable). Against the 21 *substantive* pages, density is 2.0 (14 dirty / 7 clean).
+
+> **Part 3 is the campaign's outlier — and it is a real one.** 0.40 findings/page, ~4× cleaner than the technical sections, with **all four reliability pages clean**. A clean sweep on pages seeded as high-risk is exactly where this campaign could fool itself, so the load-bearing "clean" claims were hand-verified after the run (retry config keys + defaults, telemetry enum values, the cost-tracking snippet's imports, the `deepseek-ocr` handle) — all held. **The capability/landing pages drift far less than the internals pages.**
+
+> **Stage 1 is complete: 104 pages, 1,439 claims, 109 drift findings, 0 rejected.** Density lands almost exactly at **1 finding per page**. The single strongest signal across all four parts is not *what* the page is about but *who it is for*: **pages whose author is also their reader stay current; pages that explain something to someone else drift.** Part 4 shows it cleanly — the maintainer-facing pages (drift-contracts, keyword-only, telemetry setup) came back clean, while the onboarding page, the contributor setup, the agent playbook and the docs-site notes all drifted. That is where contract budget should go.
+
+> **The zero-rejection rate is now itself a finding — and Part 4 tells us how to read it.** Across 109 confirmed findings, adversarial verify (D6) refuted *nothing*. That could mean the review agents are calibrated, or it could mean the verify pass is a rubber stamp. Part 4 supplies the tiebreaker: the post-run **false-negative** checks broke a clean verdict (Part 4's F13) that both the reviewer *and* the verifier had passed over. So the pipeline's error is not over-reporting; **it is under-reporting — false negatives, not false positives.** D6 is working; what it does not do is find what nobody looked for. The FN spot-check should be a standing step, not an afterthought.
+
+## Top defect classes so far
+
+1. **Renamed Python API/method never propagated into docs** — now the single biggest cross-part class. Building-methods: `PipelexRunner`→`PipelexMTHDSProtocol`, `bundle_uri`→`bundle_uris`, `execute_pipeline`→`execute`. Part 2 adds `OrchestratorProtocol.run`→`execute` (page-wide), `render_*_error()`→`render_inference_error()`, `model_class_from_json_schema`→`SchemaToModelFactory.make_from_json_schema`, `set_reporting_delegate`→`set_report_delegate`, `observe_*` protocol names, `LLM_MODEL_BACKEND_PAIRS`→`*_COMBOS`. Caught by an import-and-`getattr` check on doc-referenced Python names.
+2. **Doc bundle examples that fail `pipelex validate bundle`** — the dominant class in building-methods (~18 findings). Root: examples parse as TOML but fail semantic validation (missing required `description`/`domain`, inline `@` sigil, unused inputs, native-concept redeclaration, operator constraints). *(Part-2 internals pages have no bundle examples, so this class is absent there.)*
+3. **Keyword-only-convention drift in doc code snippets** (new, Part 2: F6, F9, F11, F21, F30, F32) — a doc shows a signature/call without the bare `*` the repo convention now requires; the positional form raises `TypeError`.
+4. **Parameter-table / config-model requiredness & fields** — building-methods: 5 param-table requiredness inversions vs `PipeBlueprint`. Part 2: embedded `pipelex.toml` snippet drift (F13 missing required field, F31 wrong value).
+5. **Stale enumerations** (native concepts, reasoning_effort, operator lists, deck files, img-gen handles; Part 2: flowchart directions, `EdgeKind`, fabricated `telemetry_mode` values). **Part 3 promotes this class to first rank for capability pages** — 4 of its 10 findings are a *closed* list that omits a shipped member or still names a retired one: native concepts 11-of-15, the operator roster 6-of-7 (`PipeStructure` missing, on two pages), and a `flux-2-pro` retired since v0.23.5 still listed under the Gateway. The failure mode is systematic: a list is written once, the registry grows, the list is never revisited.
+6. **Wrong defaults / behavioral claims** (dpi 150→72, PipeFunc "async required", cwd library fallback; Part 2: 0-based vs 1-based indexing, single-mutex vs double-check locking, `find_project_root` markers, output filenames, silent-no-op observer). **Part 4 extends this to repo tooling**: `align_entries` default `true`→`false` (a per-file-type override promoted to the global default), and the `.pipelex/` tree that `init` actually writes.
+7. **Repo-tooling drift** (new, Part 4 — the tail's distinctive class). Claims about `make` targets, generated file trees, `mkdocs.yml` plugins, `robots.txt` policy, and dev commands. Read the shape carefully: **not one of these was a phantom target or a missing file** — a existence check would have scored zero. The failures are *behavioral*: a target that grew a heartbeat under a stable name (`agent-test`, invalidating a whole page built on its silence), a robots.txt policy **inverted** while keeping the same variable name, a `uv` invocation that errors out, an `init` tree that grew files. **This class argues for a drift contract, not a derived check.**
+8. **The prose *around* a correct snippet** (new, Part 4 — F13, and the campaign's most uncomfortable class). The tutorial's four bundles all validate *and* dry-run clean; the sentence explaining them still teaches a false rule (`@` vs `$` distinguished by data provenance, when the real distinction is block-vs-inline rendering). A snippet validator returns green here. Only an adversarial reader catches it — and in Part 4 the *first* adversarial reader missed it too, because the false bullet sat directly above a correct, mechanically-checkable one.
+
+## Authoring-convention carve-out (Louis, 2026-07-12)
+
+A ruling that shapes both the fix triage and the check design. Omitting *scaffolding* in a focused teaching example is convention, not drift: a missing top-level `domain`, a missing pipe `description`, and an omitted `prompt` body are all things a reader fills in (with a placeholder) before running. **Still drift** (survives scaffolding): a written prompt that ignores its own declared input; wrong field names / output types / enum / default values. In building-methods this reclassified 3 findings out (all breaks-a-user: 28→25).
+
+## Drift-contract candidates (evidence for Phase 3 verdict, per D5)
+
+- **`validate-doc-bundle-examples`** (strongest for building-methods) — extract every fenced `toml`/`mthds` bundle example under `docs/`, **inject placeholder `description`/`domain` and tolerate an omitted prompt**, then run `pipelex validate bundle`. The placeholder-injection step (Louis's insight) is what makes it precise — it isolates the ~18 genuine breaks and skips every teaching fragment. Stage 0's prescreen missed all of these because it only checked TOML *parses*, not that the bundle *validates*.
+- **`doc-python-snippets-import`** (strongest for under-the-hood/advanced — Part 2 vindicates it hard) — import-and-`getattr` smoke check for Python symbols referenced in doc snippets. Catches the renamed-API class, which is the #1 defect class across both parts (F3/F8/F10/F27/F35/F37/F38 + the F30 invented param, plus building-methods' `PipelexRunner`/`bundle_uri`/`execute_pipeline`).
+- **`doc-python-signatures-vs-live`** (new, Part 2) — parse fenced `python` blocks in `docs/under-the-hood/*.md`, resolve each documented call/signature against live `inspect.signature`, flag positional uses of keyword-only params. Catches the whole keyword-only-drift class (F6/F9/F11/F21/F30/F32). Sibling to the existing `check-keyword-only` source guard.
+- **`doc-toml-config-vs-shipped`** (new, Part 2) — line-diff every fenced `toml` block labelled as `pipelex.toml` against the real shipped config (fields present + values). Catches F13 (missing required field) and F31 (wrong value); the `reasoning-controls.md` reviewer independently proposed exactly this.
+- **`operator-param-table-vs-blueprint`** — diff each operator doc's parameter table against its pydantic model's required/optional fields (catches the 5 requiredness inversions in building-methods). Needs a table parser.
+- **`doc-enumerations-vs-registry`** (new, Part 3 — the strongest candidate for the capability sections) — diff documented **closed** lists against the code's registries: `NativeConceptCode`, the `pipe_operators/` package (blueprints carrying `pipe_category="PipeOperator"`), the pipe controllers, and the shipped model deck. Catches F2/F7/F8/F9. The diff is trivial; **the hard part is deciding which lists are closed** — the exhaustive-vs-illustrative rule Part 3 had to invent would need encoding (only check lists with no "such as" / "including" / "and more" hedge), otherwise the check drowns in false positives on deliberately-partial teaching lists. **Part 4 widens the source of truth beyond code registries**: its two enumeration findings resolve against *config files* (`mkdocs.yml`'s plugin list) and the site's landing page (the `PipeStructure` omission — now confirmed on **three** pages). The single highest-value instance of this whole class: one stale operator roster, copied three times, one of them on `docs/index.md`.
+- **`make-targets-and-tooling` — a CONTRACT, not a check (new, Part 4; the clearest contract case the campaign has produced).** Part 4's distinctive class cannot be mechanized: every documented `make` target *exists*, so a target-existence script scores zero. What drifted is **behavior under a stable name** — `agent-test` grew a heartbeat (`0d2f2bf369`) ~3.5 weeks after `docs/agents/debugging-hanging-pytest-runs.md` was written on the premise that it was silent, and nothing anywhere could have flagged that. The right instrument is a drift contract (**trigger:** `Makefile`; **review targets:** the pages that describe `make` targets — `docs/agents/debugging-hanging-pytest-runs.md`, `docs/contribute/*`, `CONTRIBUTING.md`), forcing a human to look. This is the campaign's strongest evidence that some drift classes want *review obligations* rather than *scripts* — i.e. evidence for the drift-contracts design itself.
+- **⚠ A limit on `validate-doc-bundle-examples`, learned in Part 4.** The tutorial's bundles pass validation *and* dry-run, and the page still teaches a false rule in the prose explaining them (F13). **A snippet checker verifies the snippet, never the sentence beside it.** The derived-check candidates cover the *mechanical* residue of the campaign well; they do not touch the conceptual residue. Do not let a green check be read as a clean page.
+
+## Rejected / reclassified findings (audit trail)
+
+- **building-methods:** 0 rejected by adversarial verify (all 47 review-stage findings held up). 3 later reclassified as authoring convention (missing `domain`, missing `description`, omitted prompt body) → 44 counted as drift. See `building-methods.md` → "Reclassified as authoring convention".
+- **under-the-hood + advanced:** 0 rejected (all 42 held; 6 carried minor evidence corrections folded into the findings file). 0 reclassified — the D11 carve-out doesn't apply to internals/injection pages (no bundle-teaching examples).
+- **features + reliability:** 0 rejected (all 10 held). 0 reclassified — D11 doesn't apply (capability pages carry no bundle-teaching examples). The suppression risk here ran the *other* way (false negatives from the new exhaustive-vs-illustrative and floor-count rules), so the high-risk clean pages were hand-verified post-run; see the findings file's spot-check note.
+- **the tail:** 0 rejected (all 13 held). 0 reclassified. **12 findings came from the wave; the 13th came from the post-run false-negative check, which broke a page the wave had cleared** — the first time in the campaign that discipline caught something. `setup/telemetry.md` and the two contribute pages were also FN-checked and **held** (decisively).
+- **⚠ Correction to the campaign's own record (raised by Part 4).** Part 2's F24 is correct *as written* (`pipelex-agent init --config` does not accept a `telemetry_mode` field), but its **defect-class label — "fabricated `telemetry_mode` values" — is wrong**, and Part 4's seeds inherited the error. The values are real: `PostHogMode` is a live StrEnum with exactly `off`/`anonymous`/`identified` (`pipelex/system/telemetry/telemetry_config.py:27-32`), and `docs/setup/telemetry.md` documents them **correctly**. A Stage-2 fixer working from the class label alone could have gutted an accurate page. Fix F24 at its real site (`init-cli-flows.md:404`), leave `setup/telemetry.md` alone, and correct the class label in `under-the-hood-advanced.md`.
+
+## Cross-part / cross-repo carry-overs
+
+- **`docs/under-the-hood/execution-graph-tracing.md:269`** — RESOLVED as Part 2 F10 (`execute_pipeline`→`execute`, confirmed). Fold into Stage 2's fix batch.
+- **`docs/tools/cli/agent-cli.md:38`** — NEW, from Part 3 F1. The agent-CLI page documents `--mock-inputs` as just "Use mock inputs", omitting the **`(requires --dry-run)`** qualifier that `docs/tools/cli/run.md` states correctly three times. `docs/tools/cli/` is out of Stage-1 scope (D2, freshly reviewed) — but D2 reopens it precisely when a Stage 1 finding points back in, and this one does. One-line fix; fold into Stage 2.
+- **`PipeStructure` missing from the operator roster — 3 pages** (Part 3 on `features/index.md` + `features/pipe-operators.md`; Part 4 F10 on `docs/index.md`, the site's landing page). One stale closed list, three copies. Fix as a single Stage-2 batch.
+- **⚠ SPILLOVER OUTSIDE `docs/` — the stale `make agent-test` claim (Part 4 F8). NEEDS A LOUIS RULING at Checkpoint 1.** `agent-test` is no longer silent on success (it echoes a start line, a 20 s heartbeat, and a completion line — `Makefile:732-739`, `:49-65`, `:27`). The doc page F8 covers is in scope, but **the same false claim is also in `CLAUDE.md:52` ("Zero output on success") and the workspace-root `CLAUDE.md:160` ("silent on success")** — agent-facing instruction files, outside D2's `docs/` scope, and the likely *source* the doc page was written from. Fixing the page while leaving its source stale re-seeds the drift. **Question for Louis:** fold both `CLAUDE.md` fixes into the Stage-2 batch, or split them out?
+- *(No outgoing carry-overs from Part 2 — its findings are self-contained to their pages.)*
+
+## Cookbook-repo defects (handoff, not fixed here)
+
+- *(none — Stage C never ran; see the coverage disclosure below, D12)*
+
+---
+
+# Campaign close-out (Checkpoint 2, 2026-07-12)
+
+## Coverage — stated honestly
+
+Stage 1 swept **104 of the 125 in-scope hand-written pages** (denominator: `inventory.md`). The **21 `docs/cookbook/` pages were never swept** — Stage C was deferred out of the campaign entirely (D12), so no claim about cookbook-page accuracy can be made either way. Also excluded by design (D2): generated docs (`docs/errors/`, gateway models) and the freshly-reviewed sections (`docs/configuration/`, the CLI docs) — re-sweeping those would have measured nothing.
+
+## Stage 2 resolution — all four PRs merged to dev (2026-07-12)
+
+| Batch | Scope | PR → dev (squash) | Stage-1 findings | Fix-time drift (D16) | Refuted at fix time |
+|---|---|---|---|---|---|
+| S2-3 | under-the-hood + advanced | #1042 `9814e2f4c` | 42 fixed, 0 rejected | +4 (F43–F46) | 2 |
+| S2-1 | building-methods `pipes/` | #1043 `126616a88` | 30 fixed, 0 rejected | +10 (F45–F54) | 2 |
+| S2-2 | building-methods concepts + top-level | #1044 `46511ce9c` | 14 fixed, 0 rejected | +7 (F55–F61) + 1 stale pointer | 6 |
+| S2-4 | features + reliability + the tail + all cross-cutting sweeps | #1045 `cb13e5b33` | 23 fixed, 0 rejected | +10 (F11, F14–F22) | 3 |
+
+- **Every one of the 109 Stage-1 findings was fixed; 0 were rejected at fix time** — each reproduced exactly as recorded when the fixer re-verified it against the code (D6). Combined with 0 rejections at verify time, the pipeline's precision across the whole campaign is unblemished; its known error mode is false *negatives* (see the Stage-1 note above).
+- **Fix-time drift (D16) added 31 confirmed findings + 1 stale section pointer, all fixed in-batch; 13 candidates were refuted** — the refutations cluster on omission/scope/equivalence claims (a scoped list is not incomplete; an intentionally-narrower command is not wrong), the confirmations on checkable facts. Fix agents over-flag; the D6 verify gate is what keeps D16 honest.
+- **Every changed example re-validated against the real parser** (incl. dry-run execution for the load-bearing ones); blind no-context post-fix reviews ran on every batch commit — they caught 2 defects in S2-1's own new text and 3 in S2-3's (folded in pre-PR); S2-2 and S2-4 came back clean.
+- **Cross-cutting sweeps all landed:** the `PipeStructure` roster on all three pages (incl. the docs landing page), `agent-cli.md`'s `--mock-inputs` qualifier, and the D13 `agent-test` heartbeat claim everywhere — including the `pipelex/kit/agent_rules/` source templates (repo-root `CLAUDE.md`/`AGENTS.md` are generated from them) and the workspace-root `CLAUDE.md` in its own repo (commit `cc2b68c` on `feature/Follow-ups` there).
+- **Gates on the final reconciled branch:** `make docs-check` strict, `make check`, and full **`make agent-test` — all green** (the agent-test run itself displayed the start-line + 20 s heartbeats the D13 fixes now document).
+
+Counting Stage 0's mechanical fixes (dead paths/keys/links, committed at Checkpoint 0), the campaign fixed **~140 confirmed code↔doc drift defects** across the docs tree, plus the two instruction-file sweeps outside `docs/`.
+
+## Correction to the record: the observer-page "partial fix" was a stale-base artifact
+
+The tracker and S2-4's out-of-batch note recorded `docs/advanced/observer-provider-injection.md` as only partially fixed (pre-rename hook names and a broken `from pipelex import Pipelex` import "remaining"). **Verified at close-out: the page is fully fixed and byte-identical to dev.** Part 2's F38–F42 covered exactly those defects and S2-3 (PR #1042) fixed them; S2-4's session was reading a base that predated #1042's merge, so it re-reported already-fixed defects as outstanding. Close-out verification against live code: hook names match `observer_protocol.py:17-31`, payload keys match `pipe_router_protocol.py:18-47`, JSONL filenames match `LocalObserverEventType` (`local_observer.py:11-14`), the import and `Pipelex.make(...)`/`hub.set_observer(...)` calls are correct (`hub.py:263`), and `docs/advanced/index.md`'s `Pipelex.make` kwargs all exist on the real signature. **No follow-up needed.** (Lesson for future multi-PR batch campaigns: an out-of-batch "still broken" record made while sibling PRs are in flight must name the base SHA it was checked against.)
+
+## Handoff 1 — drift-contract shortlist for the Phase 3 verdict (per D5/D17)
+
+The candidates are detailed in the section above ("Drift-contract candidates"); this is the ranked verdict-ready shortlist, with the campaign's evidence base:
+
+1. **`doc-enumerations-vs-registry`** (derived check) — highest value-per-effort. Stale closed lists appeared in every part and dominate the capability pages; the diff is trivial and the sources of truth are registries/StrEnums/deck/config files. The one design cost: encode the exhaustive-vs-illustrative rule (only hedge-free lists are checked; a scope qualifier like "from PDFs and images" is part of the claim — two S2-4 refutations prove skipping scoped lists is correct, not lenient).
+2. **`validate-doc-bundle-examples`** (derived check) — kills the single biggest breaks-a-user class (~18 findings in building-methods). Needs D11 placeholder injection (`description`/`domain`, tolerated-omitted prompt) and D19 sub-pipe scaffolding for sequence sketches. **Known limit (F13): it verifies the snippet, never the sentence beside it** — a green run must not be read as a clean page.
+3. **`doc-python-snippets-import`** (derived check) — import-and-`getattr` on doc-referenced Python symbols. Catches the campaign's #1 cross-part class (renamed API never propagated into docs), which produced the worst silent failure found anywhere (the observer page's mis-named hooks silently no-op).
+4. **`make-targets-and-tooling`** (a CONTRACT, not a check) — Part 4's distinctive class is un-mechanizable: every documented target exists; what drifts is behavior under a stable name (`agent-test`'s heartbeat). Trigger: `Makefile`; review targets: the pages describing make targets (`docs/agents/debugging-hanging-pytest-runs.md`, `docs/contribute/*`, `CONTRIBUTING.md`, the kit agent-rules templates). This is the campaign's strongest evidence for the drift-contracts mechanism itself — and the finding that docs edited in the same commit as their code never drifted (`d3550330a`, `15e0a94ea`) is the mechanism observed working in the wild.
+5. Weaker / harder candidates, in descending order: `doc-python-signatures-vs-live` (keyword-only drift in snippets; needs signature resolution), `doc-toml-config-vs-shipped` (diff doc-embedded `pipelex.toml` blocks against the shipped file), `operator-param-table-vs-blueprint` (needs a Markdown table parser).
+
+**Where to spend contract budget — the campaign's headline correlation:** drift follows *audience*, not complexity. Author-is-the-reader pages (drift-contracts, keyword-only, telemetry setup, reliability) came back clean; explain-it-to-someone-else pages (onboarding, contributor setup, agent playbook, internals) drifted at up to 2.0 findings/page. Put contracts on reader-facing pages; maintainer-facing pages stay current by use.
+
+## Handoff 2 — deferred code findings (D8)
+
+**`wip/drift-hunt/deferred-code-findings.md` (kept, not pruned) carries the 10 code-side defects** the hunt surfaced but did not fix (Stage 2 was docs-only, D17). Highlights by user impact: item 3 (`GatewayTelemetryManagerInjectedError`'s remediation message names five config fields that don't exist — the one moment an error message is load-bearing, and every key it offers is wrong), item 8 (the `tomli>=2.3.0` floor admits a tomli that cannot parse documented `.mthds` deck syntax), item 9 (`pipelex build structures` can silently shadow a hand-written class after the documented migration), item 2 (cross-repo: `plxt lint` doesn't detect invalid concept references, contradicting what the docs promised). Each entry has `file:line` evidence and a candidate fix.
+
+## Lessons that outlive the campaign
+
+- **The code is the authority — even over the finding's own Reality text.** S2-3's blind review caught a Stage-1 finding whose *correction* was itself wrong (F36's fallback claim). Fixers must re-derive, not transcribe.
+- **Validate the whole snippet, not just the finding's clause.** A finding scoped to a kwarg hid a non-running example around it (S2-1's `bundle_uris` snippets, inherited from the original docs).
+- **A blind, no-context post-fix review earns its keep** — it caught defects in three of five reviewed batches (incl. Stage 1's own record), precisely because it inherits none of the batch's assumptions.
+- **The FN spot-check is a standing step.** With 0 rejections across 109 verify passes and ~140 fixes, the pipeline demonstrably under-reports rather than over-reports; the only Stage-1 defect that escaped both reviewer and verifier (F13) was caught by hand-checking a "clean" high-risk page.
+- **Run it, don't guess.** Twice, live execution reversed a wrong conclusion reached by reading (`pipelex validate --all` at Stage 0; the tomli TOML 1.1 refutation at S2-2) — and two Part-4 findings were reachable *only* by executing.
diff --git a/wip/drift-hunt/inventory.md b/wip/drift-hunt/inventory.md
new file mode 100644
index 000000000..a79280179
--- /dev/null
+++ b/wip/drift-hunt/inventory.md
@@ -0,0 +1,199 @@
+# Drift Hunt — Stage 0 inventory
+
+Every hand-written `.md` page under `docs/` in scope for the hunt, grouped by section, per decision D2 (see `TODOS.md`). Page counts per section are the denominator for the campaign's defect-density metric. Snapshot taken 2026-07-12 on `docs/Drift-hunt`.
+
+## Scope summary
+
+| Section | Pages | Notes |
+|---|---|---|
+| building-methods | 35 | MTHDS language reference (blueprint layer) |
+| cookbook | 21 | ground truth = `../pipelex-cookbook` (D3); swept in greenlight-gated Stage C, not Stage 1 (D7) |
+| features | 21 | |
+| under-the-hood | 19 | |
+| advanced | 7 | provider-injection pages |
+| reliability | 4 | |
+| contribute | 3 | 2 of 3 written/rewritten within the last days (drift-contracts, keyword-only) — expected low yield, kept per D2 |
+| get-started | 3 | |
+| (root) | 2 | `index.md`, `license.md` |
+| tools (non-CLI) | 2 | `logging.md`, `plxt.md` |
+| setup | 1 | `telemetry.md` (`gateway-models.md` excluded as generated) |
+| viewpoint | 1 | |
+| D9 additions (agents, analytics, distributed-execution, docs/CLAUDE.md) | 4 | ruled IN at Checkpoint 0 |
+| D9 additions (root `CONTRIBUTING.md`, `CODE_OF_CONDUCT.md`) | 2 | ruled IN at Checkpoint 0; live outside `docs/`, no Stage 0 seed rows |
+| **Total campaign scope** | **125** | = 104 in Stage 1 (four parts, per D10) + 21 cookbook in Stage C (D7) |
+
+## Scope-edge additions — ruled IN at Checkpoint 0 (D9, Louis 2026-07-12)
+
+Hand-written, not generated, not freshly reviewed — not named by D2's original section list; all included by ruling. They run in Stage 1 Part 4.
+
+- `docs/agents/debugging-hanging-pytest-runs.md` — agent-facing debugging guide
+- `docs/analytics/data-extraction.md`
+- `docs/distributed-execution/index.md` — the single public distributed-execution capability page
+- `docs/CLAUDE.md` — docs-site architecture notes for agents (not in the mkdocs nav, but rich in code-contradictable claims)
+- Root-level `CONTRIBUTING.md` and `CODE_OF_CONDUCT.md` — the include-targets of the docs stubs; live outside `docs/`, so the Stage 0 pre-screen (docs/-only) produced no seed rows for them.
+
+## Excluded (with reasons)
+
+- `docs/errors/**` — generated (`make gep`); fix the generator, never the pages (D2).
+- `docs/setup/gateway-models.md` — snippet-include of the generated gateway-models doc (D2).
+- `docs/configuration/**` — freshly reviewed as the config-docs drift-contract seed ack (D2).
+- `docs/tools/cli/**` — freshly reviewed as the cli-docs drift-contract seed ack (D2; `pipelex/cli/agent_cli/CLAUDE.md` likewise).
+- `docs/changelog.md`, `docs/contributing.md`, `docs/CODE_OF_CONDUCT.md` — pure snippet-include stubs (`--8<--` of root files); no content of their own.
+- `docs/overrides/**` — theme templates, not content pages.
+
+## In-scope pages by section
+
+### building-methods (35)
+
+- `docs/building-methods/adapt-to-llm-prompting-style-openai-anthropic-mistral.md`
+- `docs/building-methods/concepts/define_your_concepts.md`
+- `docs/building-methods/concepts/inline-structures.md`
+- `docs/building-methods/concepts/native-concepts.md`
+- `docs/building-methods/concepts/python-classes.md`
+- `docs/building-methods/concepts/refining-concepts.md`
+- `docs/building-methods/configure-ai-llm-to-optimize-methods.md`
+- `docs/building-methods/domain.md`
+- `docs/building-methods/kick-off-a-methods-project.md`
+- `docs/building-methods/libraries.md`
+- `docs/building-methods/packages.md`
+- `docs/building-methods/pipelex-bundle-specification.md`
+- `docs/building-methods/pipes/csv-input-and-output.md`
+- `docs/building-methods/pipes/executing-pipelines.md`
+- `docs/building-methods/pipes/index.md`
+- `docs/building-methods/pipes/pipe-controllers/PipeBatch.md`
+- `docs/building-methods/pipes/pipe-controllers/PipeCondition.md`
+- `docs/building-methods/pipes/pipe-controllers/PipeParallel.md`
+- `docs/building-methods/pipes/pipe-controllers/PipeSequence.md`
+- `docs/building-methods/pipes/pipe-controllers/index.md`
+- `docs/building-methods/pipes/pipe-operators/PipeCompose.md`
+- `docs/building-methods/pipes/pipe-operators/PipeExtract.md`
+- `docs/building-methods/pipes/pipe-operators/PipeFunc.md`
+- `docs/building-methods/pipes/pipe-operators/PipeImgGen.md`
+- `docs/building-methods/pipes/pipe-operators/PipeLLM.md`
+- `docs/building-methods/pipes/pipe-operators/PipeSearch.md`
+- `docs/building-methods/pipes/pipe-operators/PipeStructure.md`
+- `docs/building-methods/pipes/pipe-operators/index.md`
+- `docs/building-methods/pipes/pipe-output.md`
+- `docs/building-methods/pipes/provide-inputs.md`
+- `docs/building-methods/pipes/run-modes-and-backends.md`
+- `docs/building-methods/pipes/signature-pipes.md`
+- `docs/building-methods/pipes/understanding-multiplicity.md`
+- `docs/building-methods/pipes/understanding-optionality.md`
+- `docs/building-methods/pipes/working-memory.md`
+
+### cookbook (21)
+
+- `docs/cookbook/advisory-board.md`
+- `docs/cookbook/blog-article-generator.md`
+- `docs/cookbook/design-slides.md`
+- `docs/cookbook/discord-newsletter.md`
+- `docs/cookbook/extract-dpe.md`
+- `docs/cookbook/extract-gantt.md`
+- `docs/cookbook/extract-generic.md`
+- `docs/cookbook/extract-invoice.md`
+- `docs/cookbook/extract-markdown.md`
+- `docs/cookbook/extract-proof-of-purchase.md`
+- `docs/cookbook/extract-slides.md`
+- `docs/cookbook/extract-table.md`
+- `docs/cookbook/gen-expense-data.md`
+- `docs/cookbook/gen-synthetic-data.md`
+- `docs/cookbook/generate-image.md`
+- `docs/cookbook/hello-world.md`
+- `docs/cookbook/index.md`
+- `docs/cookbook/summarize.md`
+- `docs/cookbook/using-inference-plugins.md`
+- `docs/cookbook/validate-expense-data.md`
+- `docs/cookbook/write-tweet.md`
+
+### features (21)
+
+- `docs/features/advanced-customizations.md`
+- `docs/features/claude-code-skills-plugin.md`
+- `docs/features/cli.md`
+- `docs/features/cloud-storage.md`
+- `docs/features/concepts.md`
+- `docs/features/configuration.md`
+- `docs/features/cost-tracking.md`
+- `docs/features/distributed-execution.md`
+- `docs/features/document-extraction.md`
+- `docs/features/execution-graph.md`
+- `docs/features/gateway.md`
+- `docs/features/image-generation.md`
+- `docs/features/index.md`
+- `docs/features/llm-integration.md`
+- `docs/features/mthds-language.md`
+- `docs/features/pipe-operators.md`
+- `docs/features/pipeline-orchestration.md`
+- `docs/features/plxt.md`
+- `docs/features/telemetry.md`
+- `docs/features/validation-dry-run.md`
+- `docs/features/web-search.md`
+
+### under-the-hood (19)
+
+- `docs/under-the-hood/architecture-overview.md`
+- `docs/under-the-hood/build-time-elaboration.md`
+- `docs/under-the-hood/codegen-projections.md`
+- `docs/under-the-hood/distributed-content-generation.md`
+- `docs/under-the-hood/dry-run-mock-generation.md`
+- `docs/under-the-hood/error-model.md`
+- `docs/under-the-hood/execution-graph-tracing.md`
+- `docs/under-the-hood/image-handling-in-llm-prompts.md`
+- `docs/under-the-hood/index.md`
+- `docs/under-the-hood/inference-backend-plugins.md`
+- `docs/under-the-hood/init-cli-flows.md`
+- `docs/under-the-hood/orchestrator-plugins.md`
+- `docs/under-the-hood/pipe-routing-and-execution.md`
+- `docs/under-the-hood/reasoning-controls.md`
+- `docs/under-the-hood/runtime-bridge-and-transport.md`
+- `docs/under-the-hood/secrets-provider-plugins.md`
+- `docs/under-the-hood/storage-provider-plugins.md`
+- `docs/under-the-hood/stuffartefact-and-image-rendering.md`
+- `docs/under-the-hood/test-profile-configuration.md`
+
+### advanced (7)
+
+- `docs/advanced/content-generator-injection.md`
+- `docs/advanced/index.md`
+- `docs/advanced/observer-provider-injection.md`
+- `docs/advanced/pipe-router-injection.md`
+- `docs/advanced/reporting-delegate-injection.md`
+- `docs/advanced/secrets-provider-injection.md`
+- `docs/advanced/storage-provider-injection.md`
+
+### reliability (4)
+
+- `docs/reliability/automatic-retries.md`
+- `docs/reliability/durable-execution.md`
+- `docs/reliability/failure-classification.md`
+- `docs/reliability/retries-and-resilience.md`
+
+### contribute (3)
+
+- `docs/contribute/configuration-defaults-and-overrides.md`
+- `docs/contribute/drift-contracts.md`
+- `docs/contribute/keyword-only-arguments.md`
+
+### get-started (3)
+
+- `docs/get-started/build-with-claude-code.md`
+- `docs/get-started/configure-ai-providers.md`
+- `docs/get-started/mthds-language-tutorial.md`
+
+### (root) (2)
+
+- `docs/index.md`
+- `docs/license.md`
+
+### tools, non-CLI (2)
+
+- `docs/tools/logging.md`
+- `docs/tools/plxt.md`
+
+### setup (1)
+
+- `docs/setup/telemetry.md`
+
+### viewpoint (1)
+
+- `docs/viewpoint/viewpoint.md`
diff --git a/wip/kick-off-drift-hunt.md b/wip/kick-off-drift-hunt.md
index 92907875b..9da3a6925 100644
--- a/wip/kick-off-drift-hunt.md
+++ b/wip/kick-off-drift-hunt.md
@@ -1,86 +1,5 @@
-# Drift Hunt — one-off full review of the hand-written docs
+# Drift Hunt — moved
-A campaign, not a feature: sweep every hand-written doc page for code↔doc drift, fix what's stale, and come out with per-section defect-density evidence. The drift-contracts system (see [wip/drift-contracts-design.md](drift-contracts-design.md), `docs/contribute/drift-contracts.md`) enforces the floor *going forward*; this hunt backfills the floor once. It was deliberately scoped out of the tool — it runs as a project instead.
+This campaign plan was promoted to the live tracker at the repo-root `TODOS.md` on branch `docs/Drift-hunt` (2026-07-12). All state — decisions D1–D6, stage checklists, checkpoint protocol, cold-start context — lives there now; this stub only preserves the original location for links.
-## Cold-start context (update at every checkpoint)
-
-- **Status:** PLAN WRITTEN — nothing started. Next = Stage 0.
-- **Branch / worktree:** `docs/Update` in the `_docs` worktree (treat as repo root). Stage 2 fix batches may land as their own PRs — see Decision D4.
-- **Working artifacts:** everything the hunt produces (pre-screen script, suspect list, findings) lives under `wip/drift-hunt/`.
-- **Decisions taken:** D1–D6 below.
-- **Open questions:** none blocking.
-
-## Decisions
-
-- **D1 — cocode is out of scope.** Not ready for this task (user call, 2026-07-05). The whole hunt runs on the pre-screen script + Claude agent fan-outs; do not reach for cocode even for sub-steps.
-- **D2 — scope = hand-written pages under `docs/`.** Excluded as *generated*: `docs/errors/` (error reference pages, regenerated by `make gep`) and the generated gateway-models doc(s). Excluded as *freshly reviewed*: `docs/configuration/` and the CLI docs (`docs/tools/cli/` + `pipelex/cli/agent_cli/CLAUDE.md`) — they got a genuine review days ago as the drift-contracts seed acks; re-sweeping them now would measure nothing. Everything else under `docs/` is in: building-methods, features, cookbook, under-the-hood, tools (non-CLI), advanced, reliability, get-started, contribute, setup, viewpoint, index/license.
-- **D3 — cookbook ground truth is a sibling repo.** `docs/cookbook/` pages reference examples that do NOT live in this codebase — they live in `../pipelex-cookbook`. Any agent reviewing a cookbook page must verify claims against `../pipelex-cookbook` (read-only), not conclude "referenced example missing" from this tree. The pre-screen's path-existence check must special-case `docs/cookbook/` the same way.
-- **D4 — fixes land in reviewable batches.** One commit per docs section at minimum; separate PRs per section if a batch grows beyond comfortable review size. Default base: `dev`.
-- **D5 — no manifest growth during the hunt.** The drift-contracts dogfood freeze stays in force: the hunt produces *evidence* for which sections might earn a contract later; it does not add contracts to `drift.toml`.
-- **D6 — findings must survive adversarial verification before anything is fixed.** The failure mode of a big sweep is an agent confidently "correcting" a doc that was right. No finding is acted on unless a second, independent agent confirms it against the code with `file:line` evidence. Fix agents re-verify at fix time; a finding that doesn't hold up gets recorded as rejected, not silently dropped.
-
-## What "drift" means here (and what it doesn't)
-
-In scope per page: statements the code can contradict — commands, flags, config keys, defaults, file paths, API/class/function names, described behaviors and sequences, code snippets that should run or parse. Out of scope: editorial quality, tone, structure, SEO, completeness wishes ("this page should also cover X"). This is a drift hunt, not a rewrite; the standing repo rules apply to fixes (docs = current reality, no hardcoded counts, MkDocs conventions).
-
----
-
-## Stage 0 — inventory + mechanical pre-screen
-
-Deterministic checks that need no judgment, to make the semantic pass sharper and cheaper. The seed-ack dogfood showed this class is rich: several of its real catches (dead config section, phantom CLI flag, dead config field) were mechanically detectable.
-
-- [ ] Build the inventory: every in-scope `.md` page (per D2), grouped by section, committed as `wip/drift-hunt/inventory.md` with a page count per section — this is the denominator for defect density later.
-- [ ] Write the pre-screen script at `wip/drift-hunt/prescreen.py` (throwaway quality is fine — it's campaign tooling, not product; keep it self-contained, stdlib + repo venv only). Checks, each producing `(page, claim, verdict, evidence)` rows:
- - Referenced repo file paths exist in the tree (special-case `docs/cookbook/` → resolve against `../pipelex-cookbook`, per D3).
- - Mentioned config keys exist in `pipelex/pipelex.toml` / the `configs.py` model tree.
- - Mentioned `pipelex` / `pipelex-dev` CLI commands and flags exist in the actual `--help` surfaces.
- - Fenced `toml` blocks parse; fenced `python` blocks compile (`ast.parse`).
- - Internal doc links resolve (complementing, not re-implementing, the mkdocs strict build).
-- [ ] Run it; triage the raw hits into `wip/drift-hunt/suspects.md` — per section: confirmed-mechanical defects vs needs-judgment flags vs false positives (note false-positive patterns; they tune Stage 1 prompts).
-- [ ] Fix the confirmed mechanical defects immediately (they need no Stage 1) — one commit, gates: mkdocs strict build green.
-
-### CHECKPOINT 0 — STOP (suspect list in hand)
-
-Commit Stage 0 outputs + fixes; update Cold-start context. **User reviews `suspects.md` and the false-positive patterns, then decides the Stage 1 shape:** full fan-out over all in-scope sections vs worst-sections-first. Also sanity-check Stage 1's cost estimate against the observed page sizes before committing the fleet.
-
----
-
-## Stage 1 — semantic review fan-out (the judgment work)
-
-A multi-agent Workflow over the in-scope pages — requires the user's explicit go-ahead at Checkpoint 0 (fleet-scale token spend, plausibly on the order of millions of tokens; estimate concretely at Checkpoint 0).
-
-- [ ] Per page, a review agent: read the page, extract its checkable claims, verify each against the current code (cookbook pages → `../pipelex-cookbook`, per D3), report findings with severity (`breaks-a-user` / `wrong` / `misleading` / `stale-detail`) and `file:line` evidence for both the doc claim and the contradicting code. Seed each agent with the page's Stage 0 suspect rows.
-- [ ] Adversarial verify pass (D6): every finding independently re-derived by a verifier agent prompted to refute it; only confirmed findings survive. Run as a pipeline (page-review → verify per finding), not a barrier — sections should flow.
-- [ ] Aggregate into `wip/drift-hunt/findings/<section>.md` (one file per section, findings ranked by severity) plus `wip/drift-hunt/findings/SUMMARY.md`: per-section counts, defect density (findings / pages), and the rejected-findings list with reasons.
-- [ ] Log any observations relevant to the drift-contracts dogfood (e.g. "this section's drift is all mechanically detectable → derived-check candidate, not a contract") — they feed the Phase 3 verdict, per D5.
-
-### CHECKPOINT 1 — STOP (findings, no fixes yet)
-
-Commit findings; update Cold-start context. **Triage with the user:** confirm severity ranking, kill anything that smells like editorial creep (out of scope here), agree the fix order (default: by severity, then by section). This is a natural session handoff — Stage 2 needs no Stage 1 context beyond the findings files.
-
----
-
-## Stage 2 — fix in batches
-
-- [ ] Fix per section, working from `findings/<section>.md`, highest severity first. Each fix re-verified against the code at fix time (D6) — the finding is the pointer, the code is the authority.
-- [ ] One commit per section; split into separate PRs against `dev` if a batch outgrows comfortable review (D4). Gates per batch: mkdocs strict build + `make check`.
-- [ ] Note: fixing docs does NOT reopen drift contracts (docs are review targets, not triggers — only trigger files feed the ack digest). No ack churn expected; if a fix somehow touches a trigger file, resolve it with the `drift-review` skill as usual.
-- [ ] Mark resolved findings in the findings files as batches land (mark, don't delete — the record is the campaign's audit trail until the final checkpoint).
-
-### CHECKPOINT 2 — STOP (campaign close-out)
-
-- [ ] All batches landed; mkdocs strict + `make check` green on the final state.
-- [ ] Write the close-out in `findings/SUMMARY.md`: defect density per section, top defect classes, and the evidence-based shortlist of sections that might earn a drift contract (or a derived check) after the dogfood period — hand that shortlist to the drift-contracts Phase 3 verdict (TODOS.md), per D5.
-- [ ] Prune `wip/drift-hunt/`: keep SUMMARY + inventory as the record; the per-section findings files and prescreen script can go once everything is landed.
-- [ ] Update this file's Cold-start context to DONE.
-
----
-
-## Out of scope (explicit)
-
-- cocode, in any role (D1).
-- Growing `drift.toml` (D5 — evidence now, contracts later).
-- Editorial/style/structure rewrites, new content, SEO.
-- Generated docs (`docs/errors/`, gateway models) — fix the generator if one of those is wrong.
-- `docs/configuration/` + CLI docs (just reviewed via seed acks) — unless a Stage 1 finding elsewhere points back into them.
-- Other repos' docs (the hunt reads `../pipelex-cookbook` as ground truth but does not fix it; cookbook-repo defects found along the way get listed in SUMMARY.md as a handoff).
+Campaign working artifacts live under `wip/drift-hunt/`.