Factory docs, home

A record from the Factory repository, docs/claude-qualify-guard-receipt.json, shown as committed except that its local paths appear as placeholders in angle brackets, such as <local evidence archive>, standing for files that are not published. Every hash, date and count is as recorded. The build review says what the Factory's receipts support.

{
  "component": "Claude qualification script outputs (scripts/qualify-claude-producer.ts) and the Claude decision wording in src/intelligence.ts",
  "base": "1fe8a7d",
  "branch": "factory/claude-qualify-guard",
  "disposition": "implemented-awaiting-independent-review",
  "scope": "Two small defects. (1) scripts/qualify-claude-producer.ts: a Claude qualification run rewrote the byte-pinned docs/openai-producer-receipt.json at its end. The mode is now explicit: only --openai-only writes that receipt (writeOpenAiReceipt, one caller); a Claude run writes its receipt (named for the rules revision, claude-producer-r<revision>-receipt.json) and, only when accepted-scoped, its commission into <archive>/staged/ for a person to promote byte for byte, takes no --out, and writes nothing under docs/. The revision-6 branch's staging was the reference; nothing was merged from it. (2) src/intelligence.ts: the wording of the Claude facts' qualification reason and its comment now name the owner's decision of 1 October 2026 as OD-A (the Claude CLI stays unavailable for the Factory's model roles); the registry already recorded every Claude fact as unavailable under that decision, so only prose changed. Documentation: docs/agents/producer.md (command lines, outputs, test count); site/ rebuilt.",
  "producer": "Sonnet 5.5 (claude-sonnet-5-5) at high effort implemented; it is the implementer and not the verifier.",
  "policyDigests": "Unchanged: taskPolicyDigest for every task is byte-identical before and after (snapshots in <local evidence archive> and digests-after.json); prose is excluded from the digest. No transport policy changed, so CLAUDE_POLICY_DIGEST and OPENAI_POLICY_DIGEST are unchanged.",
  "untouched": "docs/claude-producer-receipt.json and docs/openai-producer-receipt.json (the byte-pinned revision-4 records), every earlier receipt, src/claude-producer.ts and src/openai-producer.ts.",
  "notEstablished": "No live qualification was run and no model was called. This is local code and documentation only; the Claude CLI stays unavailable (OD-A). Independent review has not happened.",
  "files": {
    "docs/agents/producer.md": "b45c58bbdeda03b4a3326b404c3fba9c4f9d4e585a6fceaf1535eed771481172",
    "scripts/qualify-claude-producer.ts": "192162f11b92a3df4381a220f007bfb69dc59d6b8de27492a3bba20f5c497913",
    "src/intelligence.ts": "1ea7d20523404ebff9f0a66f70ce5f1bd9d2ba4ce2b8eda4767002a2fb78047c",
    "test/qualify-claude-producer.test.ts": "fd1218e0d37003eb5c023589722b834a33dd7277c6ddbf620a747dca1dfcaf68"
  },
  "review": []
}

Source: docs/claude-qualify-guard-receipt.json