[kyle-sexton]

No related issue: decided follow-up to #10198/#10249 (LANE_CAUSES vs babysit_risk.PERSON_CODES SSOT)

Summary

clear-escalations.mjs lifts a recorded hold only when its cause is on LANE_CAUSES, an explicit list kept apart from babysit_risk.PERSON_CODES and PERSON_PREFIXES, the reasons only a person clears. Nothing checked that the two lists never overlap, so a person reason added to LANE_CAUSES.risk would have been cleared automatically.

Fix

• New lane-causes-person-codes.test.mjs reads PERSON_CODES and PERSON_PREFIXES from babysit_risk (python3 -I) and asserts for every LANE_CAUSES.risk entry: an exact entry is not a person code and does not start with a person prefix; a prefix entry (ending in ":") neither prefixes nor is prefixed by any person code or prefix.
• A separate test asserts the Python read returned non-empty sets, so a broken import cannot pass vacuously.
• The LANE_CAUSES comment now says the test pins the explicit list against the person sets (the list stays explicit, so it fails closed).

Verification

• node --test .github/scripts/pr-merge/*.test.mjs: 520 tests, 520 pass, 0 fail.
• Fail-first: adding "files-capped" to LANE_CAUSES.risk fails with lane cause files-capped is a person code; adding "high-path:" fails with lane prefix high-path: overlaps person high-path:. Both reverted.
• scripts/check-cant-fail-scan.sh against the merge base: PASS, 1 changed test file, no gating findings.
• scripts/check-changelog-fragments.sh --check-required origin/main: no fragment required.

Left out: CI lanes, which run on the PR.

Related

Follows #10198 and #10249.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-27T01:02:36.429320Z	8e610eb	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

claude-security-review reviewed 8e610eb64f55b51400d848ab1c1b3b256197e7b0: 0 finding(s), 0 as inline comments.

No security issues found.

critical-findings: 0

• .github/scripts/pr-merge/clear-escalations.mjs is a trusted path and a merge-hold control. I read the whole change and it only rewrites a comment above LANE_CAUSES. No code, list entry or constant changes, so the control does not widen or weaken. I judged that from the code, not from the PR text.
• The new lane-causes-person-codes.test.mjs only adds a check. It fails if a LANE_CAUSES.risk entry overlaps babysit_risk.PERSON_CODES or PERSON_PREFIXES. It runs python3 -I -c with fixed arguments and a path derived from import.meta.url. No PR-controlled or untrusted input reaches the subprocess or the shell. It imports only repository code, and -I isolates the interpreter.
• The test has a guard against passing vacuously: a separate test fails if the Python read returns empty code or prefix sets.
• No plugin manifests changed, so there are no manifest lines to report.
• No prepaid-spend or cost change.

[github-actions]

claude-review reviewed 8e610eb64f55b51400d848ab1c1b3b256197e7b0: 0 finding(s), 0 as inline comments.

No findings. The PR adds lane-causes-person-codes.test.mjs and updates the LANE_CAUSES comment in clear-escalations.mjs. I read both changes against babysit_risk.py.

• Person sets: PERSON_CODES is at line 1084 and PERSON_PREFIXES at line 1085, and both are non-empty at import. The test reads them correctly, and its non-empty guard stops a broken Python read from passing vacuously.
• Overlap logic: for exact LANE_CAUSES.risk entries it checks both exact membership in PERSON_CODES and the person-prefix match. For prefix entries (ending in :) it checks overlap in both directions. That matches how laneCause matches codes (code.startsWith(entry) for prefixes, === otherwise).
• Scope: the test checks only LANE_CAUSES.risk, which is correct because every person code is a risk reason. verdict entries are verify:* codes and are not person-set members.
• Comment change: it is accurate and removes the stale "to become the single source" note.
• Verification: I did not run the tests. The 520-pass and fail-first results in the PR description are the author's claims and unverified by me.
• Cost: none. The change adds a test only and spends no billed resources.
• Closing issues: the PR closes none.