[kyle-sexton]

Refs: melodic-software/claude-code-plugins#7924

Summary

Registers ci-workflows v0.49.0 intake-triage.yml (ff8cd36321423de961381e608e4a31563e74adb3) in runner policy. claude-code-plugins#7924 repins its intake caller to that SHA, and its runner-policy gate fails today with runner-target-contract: ... no reviewed runner-input contract (auto-approval declined: inputs changed since the previously reviewed ...intake-triage.yml@182aca3...).

Owner sign-off pending. The README rollout paragraph says "sign-off pending"; replace that with "signed off by hand on " when you sign it off, then mark this ready.

Fix

• components/runner-policy/policy.json: a new intake-triage.yml@ff8cd36 entry that copies the v0.45.0 entry (3c8a4b2) and adds three inputs to allowedInputs: issue-number, allowed-author-ids and allowed-bot-sender-ids. Secrets (CLAUDE_CODE_OAUTH_TOKEN) and caller permissions (contents: read, issues: write) are unchanged.
• components/runner-policy/README.md: the rollout paragraph for v0.49.0, sign-off pending.
• components/runner-policy/runner-policy.test.mjs: a test that the v0.49.0 entry is exactly the v0.45.0 entry plus those three inputs, and that it is the only v0.49.0 key.

Input diff, v0.45.0 to v0.49.0 workflow_call

Read from both files at their SHAs (ci-workflows#773 and #774 are the only intake commits in the range):

Input	v0.45.0	v0.49.0	Registered here
issue-number	absent	added	yes
allowed-author-ids	absent	added	yes
skipped-label	absent	added	no: no caller passes it (same rule v0.45.0 applied to allowed-bot-sender-ids)
allowed-bot-sender-ids	declared, unregistered	declared	yes: the #7924 caller now passes it

The other seven inputs, the outputs (triage-failed, failure-class) and the one secret are unchanged.

Security note

• Each new input is a numeric-id allowlist or an issue number. allowed-author-ids and allowed-bot-sender-ids are comma-separated numeric user ids. The job if: rejects an empty list and a leading, trailing or doubled comma, then matches ,<id>, exactly. issue-number is checked against ^[1-9][0-9]{0,9}$ in a "Resolve the issue" step, read from the API, and refused when it is a pull request.
• The OWNER or MEMBER author_association gate is replaced by the id allowlist. An opened issue runs only when its author opened it and the author's id is listed (a bot's sender id in the bot list).
• The workflow_dispatch path is admitted only for an actor whose github.actor_id is in one of the two lists. Only a listed dispatching bot's login reaches allowed_bots.
• A new skipped job (issues: write, inside the caller's grant) posts a notice and, when skipped-label is set, applies that label after checking it exists. No secret reaches it.
• No secret, runs-on routing or caller-permission change.

Verification

• Runner-policy tests: ℹ tests 314, ℹ pass 314, ℹ fail 0, including ✔ intake-triage at the v0.49.0 tag copies its v0.45.0 entry and adds only the author-gate inputs.
• node components/runner-policy/runner-policy.mjs --root . in this repository: Runner policy passed.
• Against the claude-code-plugins#7924 worktree (fix/7886-intake-author-ids, --repository-visibility public):
  • with its current vendored policy: .github/workflows/intake-triage.yml#intake-triage: runner-target-contract: the reusable workflow path@SHA has no reviewed runner-input contract (auto-approval declined: inputs changed since the previously reviewed melodic-software/ci-workflows/.github/workflows/intake-triage.yml@182aca3ad32b79fd002def94e58eb498ac49e7af)
  • with --policy set to this branch's policy.json: Runner policy passed.
  • with this branch's runner-policy.mjs: Runner policy passed.
  • That worktree was not modified (--policy flag, no file copy).
• Pre-commit hooks: typos, editorconfig, markdownlint, gitleaks and biome clean.

Related

• claude-code-plugins#7924 passes its runner-policy gate once this merges and the policy syncs to that repository.
• Precedent: feat(runner-policy): register ci-workflows v0.45.0 model-args and repin callers #797 (v0.45.0 entries).

🤖 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 2026-10-18T05:43:48.002478Z	42cfb09	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 42cfb09988cfefa7a26b68201dd9aa293e4209dc: 0 finding(s), 0 as inline comments.

No findings. I reviewed all three changed files in this PR. I did not run the test suite or the runner-policy command; the PR body's pass counts are author-claimed and unverified.

• policy.json: I diffed the new intake-triage.yml@ff8cd36… entry against the v0.45.0 entry (@3c8a4b2…) in the file. It is identical except for the three appended inputs (issue-number, allowed-author-ids, allowed-bot-sender-ids). Secrets and caller permissions are unchanged. I did not check the ci-workflows v0.49.0 workflow_call inputs themselves.
• runner-policy.test.mjs: The new constant and test are consistent with the earlier v0.45.0 and v0.48.0 tests. The test checks that the new entry equals the v0.45.0 entry plus exactly those three inputs, and that it is the only key at that SHA, so a drifting or duplicate registration would fail it.
• README.md: The added text continues the existing rollout paragraph. It matches the policy entry: skipped-label stays unregistered, with no secret or runs-on change. It records "signed off by hand on 2026-10-18".
• PR body (advisory): The body still says "Owner sign-off pending" and asks for that wording to be replaced, but the README now records the sign-off. The body is stale, so refresh it before merging. The body follows the contract (Refs: line plus Summary, Fix, Verification and Related).
• Not available: Closing issues are none (Refs: only, no Closes). No standards-index paved-path row was checked. The only path-scoped rule file, pr-body-contract.md, is body-format guidance with no paths: match to the changed files.

[github-actions]

claude-security-review reviewed 42cfb09988cfefa7a26b68201dd9aa293e4209dc: 0 finding(s), 0 as inline comments.

No security findings.

The PR registers the ci-workflows v0.49.0 intake-triage contract in components/runner-policy. The diff is policy.json, the README and the test.

• Pin: the new entry is pinned to the full 40-character SHA ff8cd36321423de961381e608e4a31563e74adb3.
• Same grant as v0.45.0: I compared it with the v0.45.0 entry. It has the same routing and runnerInput. It has the same single secret, CLAUDE_CODE_OAUTH_TOKEN, and the same caller permissions, contents: read and issues: write.
• New inputs are narrow: it adds only three inputs, issue-number, allowed-author-ids and allowed-bot-sender-ids. The README says these tighten the author gate to a numeric-id allowlist. The README also says skipped-label stays unregistered, so no caller can set it.
• Test: the new test asserts the entry equals the v0.45.0 entry plus exactly those three inputs, and that exactly one contract uses that SHA.
• Instruction surface: no standing instruction or guardrail is deleted or softened.

I read the diff and the v0.45.0 entry only. I did not check the upstream workflow at the pinned SHA, so I can't confirm the README's description of the new gate.