review-code

Reviews code files via a sandboxed Claude subprocess for bugs, security, and architecture issues.

1|Updated Feb 10, 2026
One-click install
npx skills add https://github.com/Amorim33/.agents --skill review-code-amorim33
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: review-code
Source: https://github.com/Amorim33/.agents/tree/main/skills/review-code
Command: npx skills add https://github.com/Amorim33/.agents --skill review-code-amorim33

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Getting an unbiased, critical code review is hard when you review your own work or lack a senior engineer on hand. This Skill spawns an independent Claude Code subprocess that acts as a senior staff engineer, delivering a structured review with grades, severity-tagged findings, and a ship/no-ship verdict. ## Core Features & Use Cases - Independent Subprocess Review: Runs claude -p in a sandboxed mode (--tools "") so the review is analysis-only with no code execution. - Structured Output: Returns a grade (0-5), a verdict (APPROVE, REQUEST CHANGES, or NEEDS DISCUSSION), severity-tagged findings grouped by category, and a summary of must-fix items. - Multi-File Support: Combines multiple files into a single review input, and supports resuming a review session via --resume <session-id> for follow-up questions. - Use Case: Before merging a pull request, run /review-code src/auth.ts src/api.ts to get a critical second opinion covering security vulnerabilities, error handling gaps, and performance problems. ## Quick Start Ask the AI to run the review-code skill on one or more source files, for example by typing /review-code src/app.ts.

Frequently Asked Questions about review-code

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get an AI code review of my source files?▼

Run the review-code skill with one or more file paths, such as /review-code src/app.ts. It spawns a Claude subprocess that grades the code, tags findings by severity, and returns a verdict of APPROVE, REQUEST CHANGES, or NEEDS DISCUSSION.

How to review multiple code files at once?▼

Pass multiple paths separated by spaces, like /review-code src/a.ts src/b.ts. The skill concatenates them into a temporary file with file markers, sends it to the review subprocess, then deletes the temp file after presenting results.

Can I ask follow-up questions about a code review?▼

Yes. The skill captures a session ID from the subprocess output. Run /review-code --resume <session-id> to continue the review conversation, request clarification on findings, or discuss potential fixes.

Is the code review subprocess sandboxed?▼

Yes. The subprocess runs with --tools "" which disables all tool access, making it analysis-only with no code execution. It also uses --output-format text for clean plaintext output and a 2-minute timeout.

What happens if the code review times out or fails?▼

If the subprocess returns an error or empty output, the skill shows the error and suggests retrying. On timeout, it indicates the file may be too large and recommends reviewing a smaller section of code.