Can I use this for uncommitted code?

Yes, by providing the 'staged' argument, the skill will analyze your current git staged changes before you commit them.

How does this differ from standard static analysis tools?

While static tools find syntax errors, this skill uses specialized AI agents to understand logic, architectural patterns, and business intent, providing actionable feedback on design and security.

What is the difference between Standard and Agent Team mode?

Standard mode uses parallel subagents for quick feedback on smaller tasks, while Agent Team mode creates persistent collaborators for complex reviews involving over 10 files or cross-domain dependencies.

Does it require a specific PR format?

No, the skill integrates directly with GitHub via CLI to fetch PR diffs, or it can analyze local branches and specific file paths directly.

Multi-Agent Code Review

Name: Multi-Agent Code Review
Author: rsmdt

byrsmdt

•

216

•

Collaboration Tools

Orchestrates a specialized team of AI agents to perform comprehensive, multi-perspective code reviews for pull requests and local changes.

The Multi-Agent Code Review skill transforms Claude into a sophisticated lead architect that coordinates specialized subagents to analyze code through specific lenses such as security, performance, design patterns, and accessibility. Unlike basic linting tools, this skill retrieves full file context and uses agentic workflows to identify deep architectural issues, provide actionable fix recommendations, and generate a synthesized report with a clear verdict. It is specifically designed to handle complex codebase changes by offering both parallel 'fire-and-forget' reviews and persistent 'Agent Team' modes for large-scale coordination.

Key Features

01Multi-perspective analysis covering security, performance, patterns, and testing

02Scalable execution modes including parallel subagents or persistent agent teams

03Full-file context retrieval to ensure deep understanding of code changes beyond simple diffs

04Synthesized finding reports with deduplication and prioritized severity levels

05216 GitHub stars

06Automated verdict generation (Approve/Request Changes) based on configurable severity matrices

Use Cases

01Automating consistent pull request reviews for large development teams

02Enforcing project-specific standards and 'constitutions' across complex refactors

03Conducting pre-submission security and performance audits on staged changes

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rsmdt/the-startup review

For use in Claude.ai and ChatGPT

Download Skill