A portable code-review skill for AI agents. It focuses review output on concrete findings, file references, severity, test gaps, and residual risk.
# Code Review
Use this skill when asked to review a code change, pull request, patch, or branch.
## Review stance
Lead with findings. Prioritize bugs, behavioral regressions, security issues, data loss risks, missing authorization, concurrency problems, migration hazards, and missing tests.
Do not summarize the change before findings unless there are no findings.
## Required output shape
For each finding, include:
- severity: `P0`, `P1`, `P2`, or `P3`
- exact file path and line when available
- concrete problem
- why it matters
- smallest safe fix
If there are no findings, say that clearly and name any test gaps or residual risk.