# Review

> Structured pre-commit code review in three passes: security, performance, clean code. Use when the user asks for a review, sanity check, or says /review; or before opening a PR. Runs build/tests when available. For deep security or performance use security or performance.

- Skill: `axidify/review` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add axidify/review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/axidify/review/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: Axidify (https://skillmd.com/u/axidify)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/axidify/review

---


# Review (pre-commit)

Review **before** rewriting. Only implement fixes when the user asks.

**Not this skill:** feature completeness → **audit**; deep security only → **security**; deep performance only → **performance**.

## Checklist

1. **Scope** — list changed files (session or `git diff`)
2. **Context** — run **discover-project** if stack is unclear; state unknowns briefly
3. **Diff sanity** — no unrelated files, debug logs, commented-out code, accidental edits
4. **Three passes** (quick scan; cite file + symbol for each finding):
   - **Security** — auth, validation, injection, secrets (deep list → [pass 1](references/code-review-passes.md#pass-1-security))
   - **Performance** — N+1, blocking I/O, over-fetch, payload size (deep list → [pass 2](references/code-review-passes.md#pass-2-performance))
   - **Clean code** — errors, naming, coupling, dead code ([pass 3](references/code-review-passes.md#pass-3-clean-code))
5. **Verify** — run project **build**, **lint**, and **tests** when scripts exist
6. **Report** — use [report template](references/code-review-passes.md#report-template):
   - Summary counts by severity
   - **Rewrite recommendation:** no rewrite | small patch | partial | full — and why
   - **Confidence** (high / medium / low)
   - **Verified** — commands actually run

## Do not

- Approve without applicable checks when they exist
- Refactor or rewrite unless the user asked
- Substitute for **audit** when the question is “is the feature complete?”

## Related skills

- **security** — deep security pass
- **performance** — deep performance pass
- **test** — if checks were not run yet
- **implement** — apply approved fixes
- **pr** — open PR after review passes