review-skill
Review Skill
Run specialized reviewers against a skill document to find correctness, clarity, and conformance issues.
Core Philosophy
Every issue demands a document change: real issue → fix it; false positive → add clarifying text.
Reviewers
Correctness — will this do the wrong thing?
| Reviewer | Question | Finds |
|---|---|---|
| execution | "Would this cause wrong behavior?" | Logic errors, missing steps, broken flows (all treated equally; no severity levels) |
| contradictions | "Does A contradict B?" | Conflicting instructions |
| coverage | "Is every option/branch handled?" | Unhandled branches, missing handlers |
Clarity — will this be misunderstood?
| Reviewer | Question | Finds |
|---|---|---|
| adversarial | "Where might an LLM misinterpret this?" | Fixable ambiguities, missing info |
| terminology | "Is term X used consistently?" | Naming inconsistencies |
| conciseness | "Is everything here necessary?" | Verbosity, redundancy, over-explanation |
Conformance — does this follow conventions?
| Reviewer | Question | Finds |
|---|---|---|
| checklist | "Do these specific checks pass?" | Structural issues, missing sections |
| portability | "Would this break on other LLM providers?" | Provider-specific assumptions |
Phases
@lib/... references are expanded inline by the skill loader.
@lib/001_INITIALIZE.md @lib/002_FAN_OUT.md @lib/003_COLLECT.md @lib/004_SYNTHESIZE.md @lib/005_TRIAGE.md @lib/006_HIL_PLAN_APPROVAL.md @lib/007_ADDRESS.md @lib/008_VERIFY.md @lib/009_HIL_CHANGE_CONFIRMATION.md @lib/010_STAGE.md @lib/011_COMMIT.md @lib/012_LOOP_GATE.md @lib/013_EPILOGUE.md
Quick Reference
| Phase | Purpose |
|---|---|
| Initialize | Parse args, validate target |
| Fan Out | Launch all reviewers in parallel |
| Collect | Gather and merge results |
| Synthesize | Group into themes |
| Triage | Propose fixes |
| HIL: Plan Approval | User approves proposed fixes |
| Address | Make edits |
| Verify | Confirm changes |
| HIL: Change Confirmation | User confirms edits |
| Stage | Review and stage changes |
| Commit | Create commit with proper message |
| Loop Gate | Check pass count, loop or exit |
| Epilogue | Report and end |
Flags
| Flag | Behavior |
|---|---|
--auto |
Skip HIL checkpoints (still displays plan/summary) |
-n N |
Do N passes (default: 1) |
--reviewer <category> |
Run only correctness, clarity, or conformance reviewers |
More from corygabrielsen/skills
mission-control
Coordinate complex multi-step work using task graphs and parallel background agents. Use when work requires decomposition, delegation, and long-running operations that may survive context compaction.
11loop-codex-review
Automated code review loop with progressive reasoning levels. Runs n parallel Codex reviews (configurable via -n), Claude addresses issues, climbs from low→xhigh reasoning until fixed point (all n clean). Human approval at each iteration.
11review-pr
Thorough, educational PR review process. Gathers context, explains changes, identifies issues systematically, and drafts constructive feedback collaboratively before submitting.
11orthogonalize-pr
Identify and separate orthogonal (independent) change sets within a decomposed PR. Creates branches for each set that could theoretically be reviewed or merged separately.
10decontextualize
Course-correct when artifacts overfit to conversation context. Invoke when examples, names, or language are too specific to current work.
10loop-review-skill-until-fixed-point
Iterate /review-skill on a target until fixed point. Runs review passes until all reviewers return NO ISSUES.
9