Can I review specific Git branches or commits?

Yes, the skill supports reviewing specific Git revisions, branch ranges (e.g., main..feature), and individual commit SHAs using the --rev flag or auto-detection.

Which AI models does the Code Review skill support?

The skill can dispatch tasks to various high-performance models including Claude Opus, Claude Sonnet, OpenAI GPT-5.2 Codex, and Google Gemini 3 Pro for a multi-model analysis.

Does it work with local, uncommitted changes?

Yes, if no target is specified, the skill automatically detects your staged or unstaged Git changes to provide an immediate local review.

What specifically does the 'Security' gate check for?

The Security gate analyzes the code for input validation issues, exposed secrets/credentials, SQL injection risks, and other common vulnerabilities.

Where are the review reports saved?

Spec-based reviews are saved in your project's spec directory, while general reviews are stored ephemerally in ~/.claude/reviews/ for quick reference.

Multi-Perspective Code Review

Name: Multi-Perspective Code Review
Author: srnnkls

bysrnnkls

0•

协作工具

Performs comprehensive, multi-perspective code reviews using parallel AI subagents to ensure production-grade quality and security.

The Code Review skill transforms Claude into a rigorous quality assurance lead by dispatching specialized subagents in parallel to analyze your codebase. It evaluates changes across five critical gates: Correctness, Style, Performance, Security, and Architecture. By leveraging multiple models simultaneously—including Claude Opus and GPT-5.2 Codex—it provides a diverse range of feedback that identifies logic errors, security vulnerabilities, and architectural bottlenecks before they reach production. Whether you are validating a local diff, checking a specific git revision, or ensuring a complex spec implementation is complete, this skill generates detailed, prioritized reports to streamline your development workflow.

主要功能

01Synthesized reporting that deduplicates issues and prioritizes by severity

02Parallel subagent dispatch using multiple AI models for diverse perspectives

03Smart auto-detection of Git refs, branch ranges, local paths, and specs

04Spec-compliance tracking to verify implementation against requirements

050 GitHub stars

06Automated quality gates for Security, Performance, and Architecture

使用场景

01Validating local staged changes before committing to catch edge cases

02Preparing detailed review feedback for Pull Requests to reduce manual effort

03Performing final audits of complex feature implementations against design specs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add srnnkls/tropos code-review

For use in Claude.ai and ChatGPT

Download Skill