Skip to content

Business operating system for Claude Code — 57 skills, 21 agents, smart daemon. Unified inbox (WhatsApp/Email/Slack/Telegram), autonomous PR merge, full-AWS monitoring, revenue (Stripe+RevenueCat), e-commerce (Shopify), marketing (Klaviyo/Meta/GA4), voice (Bland/ElevenLabs), APM (Datadog/NewRelic/OTEL), YOLO mode.

docs(ops-inbox): run fan-out workers on a cheaper model tier

mission-worker728e21c37ca4
diff --git a/claude-ops/skills/ops-inbox/SKILL.md b/claude-ops/skills/ops-inbox/SKILL.md
index 4c90813e..9c18f003 100644
--- a/claude-ops/skills/ops-inbox/SKILL.md
+++ b/claude-ops/skills/ops-inbox/SKILL.md
@@ -216,6 +216,7 @@ Every `/ops:ops-inbox` run should be fast for the owner, whose time is spent onl
- **Do research, context-gathering, and draft-writing in the background while the owner works.** Kick off the readers/drafters; let them build full-thread arcs, cross-channel dedup, contact profiles, and staged draft text concurrently. Surface results as they land so the owner approves in a steady stream instead of after one big serial pass.
- **Parallelism NEVER changes the safety model.** Workers are strictly READ-ONLY — they classify and return draft text only. Every outbound send stays in the main session, one draft → one `AskUserQuestion` → one approval → one send (Rule 6 + PER-DRAFT APPROVAL).
- **Respect the box concurrency ceiling** (heartbeat `MAX_BUSY`) — queue extra work rather than exceeding it.
+- **Workers run on a cheaper tier than the main session.** Pass `model` on every fan-out agent: `haiku` for retrieval, `sonnet` for reading, classifying and ordinary drafts, the top model only for legal/financial/deal threads. Never pin a model id. Change models from the next stage onward; never stop a running worker for it. Details: `references/fan-out.md`.
## Scan engine — offline script triages first, Workflow fan-out is the DEFAULT for deep per-thread work
diff --git a/claude-ops/skills/ops-inbox/references/fan-out.md b/claude-ops/skills/ops-inbox/references/fan-out.md
index a4f98a07..c653dbd8 100644
--- a/claude-ops/skills/ops-inbox/references/fan-out.md
+++ b/claude-ops/skills/ops-inbox/references/fan-out.md
@@ -49,6 +49,22 @@ volume is more than a glance.
so it can look up the same person/topic across other channels before drafting, not just
its own assigned channel's tools.
+- **Model per worker — never the session's top model for parallel workers (the
+ operator, 2026-10-01).** A fan-out inherits the main session's model by default, so
+ a run on a frontier-tier main model silently spends frontier tokens on every scanner.
+ Set `model` explicitly on each `agent()` call: `haiku` for pure retrieval (listing
+ chats, paging search results, pulling a thread), `sonnet` for reading, classifying,
+ cross-channel context and ordinary reply drafts, and the session's top model only for
+ the few threads where a wrong word costs money or goodwill (legal, financial, deal
+ terms). Sonnet is an acceptable choice for every stage when in doubt. Never pin a
+ model id — use the tier aliases so they track the current generation (`ops-rules`
+ Rule 17).
+- **Change models from the next stage onward, never mid-flight.** A running worker is
+ sunk cost; stopping it discards its work (the resume cache only reuses agents that
+ finished with an unchanged prompt and options, and a changed `model` is a changed
+ option). Read `journal.jsonl` first: if no agent has completed, a stop gains nothing.
+ Apply a model change to stages that have not started, or to the next run.
+
**Canonical scan workflow.** Pass the available channels in via `args` (the orchestrator
builds the list from the detected-available channels), so the script body stays stable:
@@ -138,7 +154,7 @@ const scans = (await parallel(CHANNELS.map(c => () =>
\`for each (needed later to reply). Cover ~last 7 days plus \` +
\`anything clearly still open. Retry the documented reconnect handshake before reporting \` +
\`reachable=false. Never fabricate conversations.\`,
- { label: \`scan:\${c.key}\`, phase: 'Scan', schema: SCAN_SCHEMA }
+ { label: \`scan:\${c.key}\`, phase: 'Scan', schema: SCAN_SCHEMA, model: 'sonnet' }
)
))).filter(Boolean)
@@ -150,7 +166,7 @@ return await agent(
\`Return ONLY structured JSON with buckets: needsReply[], waiting[], fyi[], unreachable[]. \` +
\`Each item: {channel, who, summary, chatId, lastMessageAt}. Order needsReply most-urgent \` +
\`first. Do NOT draft replies — that happens in the main session under the per-message gate.\`,
- { label: 'synthesize', phase: 'Synthesize',
+ { label: 'synthesize', phase: 'Synthesize', model: 'sonnet',
schema: { type: 'object', additionalProperties: true } }
)
`,