eigent-ai/review

Review

Overview

Use this skill for code-review work. Prioritize bugs, behavioral regressions, security risks, performance issues, and missing tests. Findings come first, ordered by severity, with concrete file and line references whenever possible.

Workflow

1. Inspect the requested diff, files, branch, or pull request context.
2. Review across five axes:
   • Correctness.
   • Security.
   • Performance.
   • Readability.
   • Maintainability.
3. Check whether tests cover the changed behavior and important failure modes.
4. Keep feedback specific and actionable; avoid broad style commentary unless it affects maintainability or correctness.
5. Flag issues by severity:
   • Blocking: should not ship.
   • High: likely bug, security issue, or serious regression.
   • Medium: meaningful risk or missing coverage.
   • Low: small maintainability or clarity improvement.
6. If no issues are found, say so clearly and mention residual test gaps or assumptions.

Output Pattern

1. Findings first, ordered by severity.
2. Open questions or assumptions.
3. Brief change summary only after findings.
4. Tests or verification reviewed.

Example Prompts

• /review - Review this pull request. Flag anything that affects security, performance, or will be hard to maintain in 6 months. Be direct about blocking issues.
• /review - Audit this API endpoint handler for security issues. Check for injection risks, missing auth checks, improper error handling, and data exposure.
• /review - This is going to production tomorrow. Do a final review focused only on breaking changes, regressions, and anything that could cause an incident.