[kyle-sexton]

Refs: melodic-software/claude-code-plugins#8491

Summary

Registers ci-workflows v0.60.0
(82ca5ca9be6e18a1addb7986eae15689339aedbc) contracts for intake-triage.yml,
pr-review.yml and pr-review-security.yml, so claude-code-plugins can repin
intake triage from v0.58.0 and both review lanes from v0.59.1 to v0.60.0. The
intake-triage repin carries ci-workflows#854,
which claude-code-plugins#8491 needs.

Review: the GitHub Security & Cost session reviews this PR. The gate cannot
auto-approve these entries, so the copies rest on the review below. The README
rollout line says "sign-off pending"; replace it with "signed off by hand on
" at sign-off.

Fix

• components/runner-policy/policy.json: verbatim copies at 82ca5ca9 of the
  v0.58.0 (5e18f3d6) intake-triage entry and the v0.59.1 (8bd4d7f9)
  pr-review and pr-review-security entries. No input, secret mapping, caller
  permission or routing is added or widened.
• components/runner-policy/README.md: the v0.60.0 rollout line, sign-off
  pending.
• components/runner-policy/runner-policy.test.mjs: a test that the three
  v0.60.0 entries equal their previous entries and are the only keys at that SHA.

Review basis

Git blob SHAs of the workflow files and tree SHAs of the commit-relative ($/)
actions they use, from the contents API at each ref:

Path	5e18f3d6 (v0.58.0)	8bd4d7f9 (v0.59.1)	82ca5ca9 (v0.60.0)
workflows/intake-triage.yml	8ff8eac3	8ff8eac3	0c69f08c
workflows/pr-review.yml	47a2b1c9	4f3d4e44	4f3d4e44
workflows/pr-review-security.yml	16e9be5b	30b595b9	30b595b9
actions/check-subprocess-sandbox	d9d60153	d9d60153	92d69c1b
actions/report-lane-outcome	5f583be1	5f583be1	0b6bb70a
actions/run-claude-cli	0edc45c1	0edc45c1	6234a8b1
actions/apply-claude-managed-settings	1301e911	1301e911	1301e911
actions/check-author-trust	3517c76f	3517c76f	3517c76f

1. intake-triage.yml (v0.58.0..v0.60.0, four hunks per job pair, none in
   on.workflow_call, permissions or runs-on): the issue file gives each
   comment's html_url and the prompt says the model reads it
   (ci-workflows#854);
   the report step passes lane-id: intake-triage to report-lane-outcome.
2. report-lane-outcome (ci-workflows#853):
   new optional lane-id and handoff inputs; with a lane id and a credential
   secret name that maps to a kind, each annotation also carries a
   lane-outcome {json} record built from the same closed-pattern facts as the
   existing tokens. The credential appears as its secret name only, never a value.
3. check-subprocess-sandbox and run-claude-cli/lane.cjs
   (ci-workflows#849):
   the sandbox probe's error text names Docker's seccomp profile. Message text
   only.

The review workflow files are byte-identical to v0.59.1; their change is the
commit-relative actions above, which is the diff the gate declines to
auto-approve.

Verification

• npm run test:runner-policy: 327 pass, 0 fail, including ✔ the intake-triage and review contracts at the v0.60.0 tag copy their previous entries forward verbatim.
• npm run lint:runner-policy: Runner policy passed.
• In the claude-code-plugins repin worktree, node .github/standards/runner-policy/runner-policy.mjs --root .:
  • with its vendored policy (identical to standards main): three
    runner-target-contract failures, auto-approval declined for
    intake-triage.yml, pr-review.yml and pr-review-security.yml at
    82ca5ca9;
  • with this branch's policy.json copied in: Runner policy passed. (file
    restored afterwards).
• lefthook pre-commit (typos, editorconfig, markdownlint, gitleaks, biome): clean.

Related

• The claude-code-plugins repin PR passes its runner-policy gate once this
  merges and syncs.
• ci-workflows v0.60.0.
• Precedent: #838 (v0.59.1 review contracts).

🤖 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 2026-10-22T13:27:38.291723Z	9cc1640	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-review reviewed 9cc1640f96fd2662b92199920d23b628afe8564e: 0 finding(s), 0 as inline comments.

No findings. I reviewed all three changed files and posted no inline comments.

What I checked

• Copies: I read policy.json and compared each new v0.60.0 entry with its source. The intake-triage entry at 82ca5ca9 matches the v0.58.0 (5e18f3d6) entry on routing, role, inputs, secrets and caller permissions. I did not open the v0.59.1 pr-review and pr-review-security source entries myself; the new runner-policy.test.mjs test asserts they match.
• Test: It checks that the three new entries equal their previous entries, and that exactly these three keys exist at the v0.60.0 SHA.
• README: The new rollout paragraph sits after the v0.59.1 paragraph. Its SHA, dates and ci-workflows PR references agree with policy.json and the PR context.
• Closing issue: There is none. The PR uses Refs: melodic-software/claude-code-plugins#8491, so there are no acceptance criteria to check.
• Path-scoped rules: The only rule, pr-body-contract.md, covers the PR body. The body opens with a Refs: line and has Summary, Fix, Verification and Related sections.
• Plugin manifests: The diff changes none.
• Paved paths: I did not resolve the standards index, so I checked nothing about paved paths.

What I did not verify

• I did not run the tests or the lint. The PR's claims of 327 passing tests and Runner policy passed. are the author's and are unverified.
• I did not compare the blob and tree SHAs in the PR's "Review basis" table against ci-workflows. That table is also the author's.

Observation, not a finding

• The README line for v0.58.0 still says "sign-off pending", and this PR copies that entry forward. That line is on untouched lines.

[github-actions]

claude-security-review reviewed 9cc1640f96fd2662b92199920d23b628afe8564e: 0 finding(s), 0 as inline comments.

Security review: PR #842 at 9cc1640

Result: no security findings. The PR is open and the diff has security-relevant surface, so I read all three files.

What changed. The PR registers three reusable-workflow contracts in components/runner-policy/policy.json at ci-workflows SHA 82ca5ca9be6e18a1addb7986eae15689339aedbc (the v0.60.0 tag): intake-triage, pr-review and pr-review-security. It also adds a README rollout note and a test.

Why it is safe.

• Verbatim copies. I compared each new entry against its predecessor by hand. intake-triage matches the 5e18f3d entry, and pr-review and pr-review-security match the 8bd4d7f entries. For each, routing, runnerInput, role, allowedInputs, allowedSecrets and allowedCallerPermissions are identical.
• No widening. The new entries add no input, secret mapping or caller permission. id-token: write and the secret names are carried over unchanged.
• Test guard. The new test in runner-policy.test.mjs asserts that each v0.60.0 entry deep-equals its predecessor. It also asserts that exactly three keys carry the new SHA, so a stray extra approval would fail.
• Pinning. The new keys are full 40-character SHA pins, not tags or branches.
• No instruction-surface deletions. The diff removes nothing, so the instruction-exception register was not needed. No plugin manifests are touched, so there are no manifest <field> lines.

Not verified. The README says the upstream v0.60.0 workflow files keep the same interface. I could not confirm that against the ci-workflows repo, because the gh api call to the tag was blocked in this environment. The in-repo gates, including the commit-relative actions diff that the README says needs manual sign-off, are the control for that claim. It is not something this diff gets wrong.