How does it handle security vulnerabilities?

It leverages the built-in /security-review skill to perform deep audits and maps findings to priority levels (P1-P3) based on severity, such as Critical or High risks.

Can it automatically fix the issues it finds?

Yes. After displaying the integrated report, the skill provides an interactive prompt allowing you to fix all issues or select specific ones to be applied via the Edit tool.

What exactly does this skill review in my code?

It reviews local git changes (staged and unstaged) across four main agents: Claude Quality (8 dimensions), Security (vulnerabilities), Codex (logical analysis), and Simplify (complexity reduction).

Where are the review reports stored?

Reports are automatically saved as timestamped markdown files in the .claude/tmp/ directory within your project's root folder for future reference.

Do I need to have Codex installed to use this skill?

The skill attempts to use the Codex CLI if available for a secondary perspective, but it is designed to fail gracefully and continue with Claude-based reviews if Codex is missing.

Integrated Multi-Agent Code Review

Name: Integrated Multi-Agent Code Review
Author: shiiman

byshiiman

•

セキュリティとテスト

Conducts comprehensive, parallel code reviews across four specialized dimensions including security, quality, and technical simplicity.

This skill streamlines the code review process by orchestrating four specialized agents—Claude, Codex, Security, and Simplify—to analyze local git changes simultaneously. It evaluates code across eight deep-dive quality dimensions, such as performance, design patterns, and error handling, while performing security audits using built-in tools. Results are integrated into a single, prioritized markdown report with actionable feedback, allowing developers to identify critical vulnerabilities and architectural issues before they ever reach a pull request.

主な機能

01Interactive fix application to automatically resolve detected issues in the codebase

02Comprehensive Git diff analysis for both staged and unstaged local changes

033 GitHub stars

04Automated markdown report generation with P1-P3 priority mapping and IDs

058-point deep quality analysis covering design, performance, tests, and maintainability

06Parallel 4-agent review (Claude, Codex, Security, and Simplify) for maximum coverage

ユースケース

01Pre-commit self-reviews to catch high-severity security vulnerabilities and logical errors

02Generating detailed, timestamped documentation of code changes and their technical impacts

03Automated quality gating to ensure adherence to DRY principles and design best practices

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add shiiman/claude-code-plugins review

For use in Claude.ai and ChatGPT