By name: subagent-review license: MIT description: "USE FOR: Guardian and critic peer reviews using five native reviewer perspectives, with guardian aggregation. DO NOT USE FOR: implementation, dispatcher fan-out, model/tier overrides, or public posting."
Subagent Review
UTILITY SKILL: Use the current runtime's five native reviewer perspectives for guardian/critic review. Do not implement, delegate approval, dispatch fan-out, override models/tiers, or build cross-engine reviewer pools.
Normal message lane: orchestrator -> guardian -> critic -> guardian -> orchestrator.
Guardian and critic are peer review roles. Guardian aggregates the final result.
Default set: security, architecture, historian, code, and QA.
1. Boundary
This skill owns reviewer perspectives, guardian/critic review flow, synthesis,
severity, and output expectations. postman.md owns live route, reply flow,
verdicts, no-bypass gates, and public-posting approval gates. github owns
GitHub mechanics; create-review-comment owns user-facing PR comment drafts.
2. Workflow
- Run the cheapest relevant verifier first.
- For substantive reviews, use the five defaults before synthesizing. Direct review is acceptable for trivial follow-ups if stated.
- Run each wave in parallel; close it before second-wave validators.
- Add data or research reviewers only for specialized questions; do not replace the defaults.
- Use native subagents only. No dispatcher, model/tier overrides, or
cross-engine pool. Per-agent defaults live in
subagents/metadata.nix. - Give each subagent bounded read-only paths, context, and output shape.
- Suppress unverified findings, deduplicate, and produce the verdict or recommendation.
3. Role Rules
- Guardian uses Codex reviewers and aggregates the final verdict.
- Critic uses Claude native reviewers and returns independent review evidence.
- Critic sends the review result to guardian for aggregation, not orchestrator.
- Reviewer subagents must not edit, commit, push, or approve.
- If evidence is incomplete, continue direct review.
4. Output
Preserve perspective, paths, checked evidence, severity, confidence, and gaps. Final verdicts use the guardian-mediated aggregation surface, not raw subagent output.