# Code Reviewer

> Production-grade structural code review for correctness, security, performance, and release risk.

- Skill: `eddieran/code-reviewer` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add eddieran/code-reviewer`
- Raw SKILL.md: https://api.skillmd.com/api/skills/eddieran/code-reviewer/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: eddieran (https://skillmd.com/u/eddieran)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/eddieran/code-reviewer

---


# Code Reviewer

Use this skill when the user wants a real engineering review instead of lint-style commentary.

## Agent Compatibility

- Claude Code: supported
- Codex: supported

## Inputs

- A diff, commit range, pull request, or a focused set of changed files
- Any failing test output, benchmark output, or bug report that motivated the review

## Workflow

1. Determine the exact review scope before commenting.
2. Read surrounding code, not just the modified lines, until the behavior is clear.
3. Prioritize findings by user impact:
   - correctness and data loss
   - security and trust boundaries
   - concurrency and state corruption
   - performance regressions
   - migration and rollout risk
4. Call out missing coverage whenever behavior changed without tests.
5. Distinguish between confirmed defects, plausible risks, and open questions.

## Output Contract

- Findings first, ordered by severity.
- Each finding includes:
  - a short title
  - affected file paths
  - the concrete failure mode
  - why the issue matters in production
  - the smallest safe fix
- If no material findings remain, say so explicitly and mention residual testing gaps.

## Guardrails

- Do not flood the user with style nits.
- Do not speculate without a code path or failure scenario.
- Do not propose broad rewrites unless the current design is unsound.