[purp]

Summary

Prepare acceptance-triggered labeling dispatch and cross-host item coordination. The workflow is dormant by default and uses trusted default-branch code, a shared-runner adapter, explicit application gating, and the existing reviewer reconciler.

This integration draft is based on #5407, whose commits are currently included in the diff against main. Rebase onto the PR for #5411 when available, then complete the actual helper/writer wiring and observed end-to-end validation. Automatic execution remains disabled.

Related Issue

Closes #5412.

The authenticated author has verified ADMIN access to both repositories. #5412 is accepted.

Changes

• Add dormant acceptance-label, authorized human command/force, and manual dispatch triggers, including the restricted GITHUB_TOKEN acceptance-writer bot path. Creation/reopening and bookkeeping labels do not select classification.
• Add the acceptance-writer dispatch hook and provisional feat(review): connect the end-to-end labeling runner #5411 runner adapter without duplicating eligibility, classification, checkpoint or mutation logic.
• Add a GitHub-ref mutex with unique owner audit records, ownership checks, and conservative manual recovery for failed or uncertain operations.
• Preserve recovery reports before downstream failures and hand successful/unchanged PR application to the existing reviewer reconciler.
• Register offline tests and document preview/application controls and the exact remaining integration work.

Before readiness:

• Rebase onto feat(review): connect the end-to-end labeling runner #5411, use its real runner interface/statuses, and wire every mutation entry point to the shared mutex without nested locking.
• Connect feat(review): automatically label new PRs for accepted work #5260's actual successful acceptance write to dispatch before draft returns or reviewer routing, and test writer failure handoffs.
• Obtain and implement the human-approved policy for materializing implicit issue acceptance through shared accepted-work policy.
• Validate actual helper/runner eligibility, duplicate/checkpoint, uncertain-write, serialization and reviewer recovery paths.
• Verify credentials, custom Git refs, dedicated CMA setup and observed previews/one-item application before explicit activation.

Release Notes

Testing

• 43 offline labeling tests pass with uv run --no-project --python /usr/bin/python3 --offline python -m unittest discover -s tasks/scripts -p '*label*test.py'.
• 30 existing reviewer reconciliation tests pass.
• Python syntax, TOML task registration, local documentation links, and staged whitespace checks pass.
• Independent code review completed; fixed report loss during downstream failure and missing owner-execution metadata.
• Ruff and Actionlint: unavailable locally; tool downloads were blocked by network access. Markdownlint was not run.
• Actual feat(review): connect the end-to-end labeling runner #5411/feat(skills): extract issue and PR labeling #5355/feat(review): automatically label new PRs for accepted work #5260 integration and live end-to-end runs remain pending. Tests use an injected runner/fake GitHub service; no model sessions, live ref writes or automatic labeling were performed.

Checklist

• Follows Conventional Commits.
• Commits are signed off (DCO).
• Automatic preview and application remain disabled.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.