code-review

Files (1)

• SKILL.md

SKILL.md (first lines)

# Code Review

## Review Stance

Find defects, not trivia. Findings come first.

## Workflow

1. Inspect the diff and enough surrounding code to understand behavior.
2. Identify changed contracts: API, data, auth, UI workflow, persistence, deployment, or public types.
3. Look for correctness, data integrity, security, compatibility, performance, and test gaps.
4. Check whether docs, `.ai_project.md`, or evolution records should change.
5. Verify line references are tight and actionable.
6. If no findings are found, say so and mention residual risk or test gaps.

## Severity

- `P0`: production breakage, data loss, critical security, impossible rollback.
- `P1`: likely user-visible bug, security flaw, major regression, API/data contract break.
- `P2`: credible edge-case bug, missing important test, maintainability risk with concrete impact.
- `P3`: minor issue, small cleanup, low-risk polish.

## Review Checklist

- Does behavior match the request?
- Are boundary conditions handled?
- Are errors surfaced and logged usefully?
- Are auth and permissions enforced server-side where needed?
- Are database constraints, migrations, and indexes safe?
- Are API status codes and schemas compatible?
- Are UI loading/empty/error states complete?
- Are tests covering the changed behavior?
- Are project docs updated when long-term context changed?

## Output

Use findings first, then open questions, then brief summary:

```markdown
发现：
- [P1] ...

开放问题：
- ...

残余风险：
- ...
```

Include file and line for each finding. Do not pad with style nits.

From PAISHU-AI/PAI-zong-NB-v5.8, MIT.

More from PAISHU-AI/PAI-zong-NB-v5.8