Part of the experimental Factory suite of builder-skills. Reads .agent-factory/config.yaml (artifact-autonomy-policy) and its own optional skill_prompts overlay, which adds project guidance but cannot grant permission. It reviews a filtered queue of PRs. For one long-lived PR the Factory uses builder-factory-babysit-pr instead.
For each candidate: read live state from the configured host, excluding drafts, PRs outside the filters, and PRs that already have a current review unless policy asks for re-review. Inspect the diff, linked issues, required checks, review threads, author eligibility, and the exact head revision. “Treat bot findings as leads and preserve human review direction unless source evidence disproves it.” Report actionable findings with file, line, impact, and a concrete fix. “If none exist, record that outcome without inventing a comment.”
Review is one of four separately gated actions. The others are reply or other PR write, approve, and merge. “Host-level mergeability, one green check, or a bot approval does not prove all gates passed. Re-read the exact head and live state immediately before an approval or merge. Restart a configured soak if the head or a gate changes.” Unavailable or partial state counts as unknown, never as clean.
In the wiki’s stage-review code-review cluster, this is the only member designed for a queue of other people’s PRs on a schedule. The others review the agent’s own diff.
See Also
- builder-factory-babysit-pr — the one-PR counterpart.
- ce-code-review · gstack-review · mp-code-review — the code-review cluster.
- stage-review — the canonical stage this implements.