memories: document the Mechanism: discharge added by #4410

[d-morrison]

Before: memories/claude-code-hooks.md said no-mistake-without-a-hook.py has two escapes, HOOK_WORK and NOT_HOOKABLE, and that a mechanism already built elsewhere "matches none of them". #4410 (merged as 1cc4c27) added a third, MECHANISM_CITED, so that sentence no longer describes the hook. #4410's review raised this as a non-blocking note.

After: the entry describes the Mechanism: <GitHub issue or PR URL> line form, says which shapes do not discharge it (word mid-sentence, URL on another line, non-GitHub URL), and adds a Do bullet for it.

How: one hand edit, checked against the regex and test cases in hooks/no-mistake-without-a-hook.py and hooks/test-no-mistake-without-a-hook.py on main. The file is 1194 lines, under the 1250 cap.

Refs #2389

🤖 Generated with Claude Code

https://claude.ai/code/session_01Vwcai8NPoZC3GjwGwhcKHa

---

Generated by Claude Code

[d-morrison]

/review

---

Generated by Claude Code

[github-actions]

👀 Claude Review Dispatched — run reviewing PR #4411 at e38f90b.

[github-actions]

Claude finished review — View run

---

Review: PR #4411 — "memories: document the Mechanism: discharge added by #4410"

Scope: A single, minimal documentation fix to memories/claude-code-hooks.md, closing a gap flagged as a non-blocking note in PR #4410's review. No code changes.

What I checked

• Accuracy against the implementing code. The new prose describes MECHANISM_CITED in hooks/no-mistake-without-a-hook.py:125-129. I compared each claim to the actual regex:
  • "a line that starts with the label Mechanism:" — matches ^[ \t]*(?:[-*+][ \t]+)?(?:\*\*|__)?Mechanism(?:\*\*|__)?[ \t]*:.
  • "an optional list bullet and bold markers are allowed" — matches the (?:[-*+][ \t]+)? and (?:\*\*|__)? groups exactly.
  • "followed on that same line by a GitHub issue or PR URL" — matches [^\n]*?https://github\.com/[\w.-]+/[\w.-]+/(?:issues|pull)/\d+.
  • "The word mid-sentence, the URL on a different line, or a non-GitHub URL does not discharge it" — correct: the ^ anchor (with re.M) rejects mid-sentence occurrences, [^\n]*? can't cross a newline, and the literal https://github\.com/ requires that host.
  • All three claims are exact, not approximate paraphrases.
• Cross-references resolve. ai-config#4410 is a real, merged PR ("no-mistake-without-a-hook: discharge on a "Mechanism:" citation", merged 2026-10-08T13:33:53Z) and ai-config#2389 is a real, closed issue matching the described scenario (cross-repo/subagent mechanism with no hooks/ path in the transcript). No hallucinated references.
• No other docs needed the same update. Grepped the repo for NOT_HOOKABLE/MECHANISM_CITED outside this file — no other Markdown doc describes this hook's escapes, so there's nothing else left stale.
• Line-count claim. PR body states the file is "1194 lines, under the 1250 cap"; actual count at HEAD is 1193 — a trivial one-line discrepancy (likely counted before/after a trailing-newline nuance), immaterial to the real constraint (the documented 1250-line CI gate per memories/mistake-patterns.cases.md and others), and well clear of it either way.
• Style/formatting. New lines are plain ASCII, one sentence per line (matches the file's existing semantic-line-break convention), and the new Do bullet follows the existing Do/Don't list format and pattern/anti-pattern pairing convention from CLAUDE.md. No raw HTML, no fabricated API/function references.
• No prior review rounds to reconcile — this is the PR's only commit and I found no earlier claude[bot] review comment thread on it to check for already-addressed findings.

I found no bugs, no CLAUDE.md violations, and no hallucinated content. This is a correct, well-scoped, self-contained doc fix with no blocking issues.

Verdict

Ready for merge

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "e38f90b81be0fc1523cb9a048845d347f9c63576",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "memories/claude-code-hooks.md's new MECHANISM_CITED paragraph was checked line-by-line against the regex in hooks/no-mistake-without-a-hook.py and matches exactly, with no unsupported claims.",
  "holistic_assessment": "Scope is a single documentation correction closing a gap from PR 4410's review; references to ai-config#4410 and ai-config#2389 resolve correctly and no other file needed the same update."
}

Reviewed commit: e38f90b

[github-actions]

💰 Cost: $0.6394 (review) — run