[kyle-sexton]

Refs: melodic-software/claude-code-plugins#9276

Summary

Registers pr-review-cli.yml@4edeccd87cf60ef539ce5109b78454217d5f2049
(ci-workflows v0.64.0)
so callers can pin it and pass its two new optional inputs, hold-label and
hold-paths.

Fix

v0.64.0 adds an optional hold step to the status job
(ci-workflows#864).
The new contract copies the v0.63.1 entry forward and appends hold-label and
hold-paths to allowedInputs. Routing (hosted-only), fixedRunsOn
(ubuntu-24.04) and the two reviews-role secrets are unchanged. No caller
permission is added: the hold step writes through the status job's existing
pull-requests: write grant. A diff of the workflow between v0.63.1 and v0.64.0
shows no other input, secret or runs-on change.

Verification

• node --test runner-policy.test.mjs in components/runner-policy: 338 pass,
  0 fail, including the new test that the v0.64.0 entry equals v0.63.1 plus the
  two hold inputs.
• npm run lint:runner-policy: passed.
• node --test components/claude-lanes/repin-policy-lockstep.test.mjs: 23 pass.
• Pre-commit (typos, editorconfig, markdownlint, gitleaks, biome): clean.

Related

• Same shape as feat(runner-policy): register the ci-workflows v0.63.1 pr-review-cli contract #849 (v0.63.1) and feat(runner-policy): register the ci-workflows v0.63.0 pr-review-cli contract #848 (v0.63.0).
• After merge, the push-triggered maintenance-sync-standards run opens the
  sync PR into claude-code-plugins' .github/standards/runner-policy/policy.json.

🤖 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 2026-10-23T05:19:36.942965Z	3961ac9	PR opened

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 3961ac99250bf1b06edb55374bedf948cd913e57: 0 finding(s), 0 as inline comments.

Security review: PR 850 (head 3961ac9)

No security findings.

The PR adds one pinned reusable-workflow contract for pr-review-cli.yml@4edeccd… to components/runner-policy/policy.json, plus README text and a test. I read the full diff.

• Copy-forward: The entry matches the v0.63.1 contract (hosted-only, fixedRunsOn: ["ubuntu-24.04"], the same two allowedSecrets). The only change is two new input names, hold-label and hold-paths. The new test in runner-policy.test.mjs asserts this by deep-equality against the v0.63.1 entry.
• No widening: No secret, permission or routing change. The README says the status job's hold-label and marker comment use its existing pull-requests: write grant, so no caller permission grows.
• Pin shape: The contract is keyed by a full 40-character commit SHA, not a mutable tag, so a moved tag can't change what runs.
• Trust boundary: The policy allowlists input names only. How hold-label and hold-paths are handled inside ci-workflows lives outside this diff.
• Not verified: My gh api calls were denied approval. I did not confirm that the SHA is the v0.64.0 tag, or that the workflow at that SHA takes only these inputs and secrets. The reviewer should check both.
• Skips: The instruction-surface lens and plugin-manifest check don't apply, because no instruction file, hook or plugin.json / marketplace.json changed. Actions hardening is zizmor's lane.

manifest: no manifest fields were touched.

critical-findings: 0

[github-actions]

claude-review reviewed 3961ac99250bf1b06edb55374bedf948cd913e57: 0 finding(s), 0 as inline comments.

No findings. I reported nothing inline.

What I checked (read the diff and the repo files myself):

• Policy entry: the new v0.64.0 pr-review-cli.yml@4edeccd… entry in components/runner-policy/policy.json matches the v0.63.1 entry (83b739f…) on routing (hosted-only), fixedRunsOn (ubuntu-24.04) and both reviews-role secrets. Its allowedInputs are the v0.63.1 list plus hold-label and hold-paths, which is what the PR says.
• Test: the new runner-policy.test.mjs case builds the expected entry from the v0.63.1 entry, so it fails if the two drift. It also asserts exactly one contract per SHA. It follows the v0.63.1 test's shape.
• README: the added paragraph in components/runner-policy/README.md continues the v0.63.0 / v0.63.1 sequence in the same style. It matches the policy entry.
• Rules: .claude/rules/pr-body-contract.md has no paths: scope. The PR body opens with Refs: and has Summary, Fix, Verification and Related sections, so it meets the body contract. There is no closing issue, so there were no acceptance criteria to check.

Not verified: I could not run node --test here because the sandbox required approval. The PR's reported results (338 runner-policy tests, 23 lockstep tests, lint and pre-commit clean) are the author's claims and I have not confirmed them. I also could not check the author's claim that the v0.63.1 to v0.64.0 workflow diff has no other input, secret or runs-on change. I have not read the ci-workflows source.