How does the multi-agent review system work?

The skill launches three specialized subagents—Bug Hunter, Security, and Codebase Fit—which analyze the code simultaneously from different perspectives to ensure comprehensive coverage.

Can I review code that hasn't been committed yet?

Yes, the skill automatically checks for uncommitted and staged changes using git status and git diff before looking at commits or PRs.

Can the skill fix the bugs it finds?

Yes, after presenting the review summary, the skill offers a prompt to run a /fix command, which uses engineering agents to automatically resolve critical and warning issues.

Does it support reviewing GitHub Pull Requests?

Absolutely. By providing a PR number, the skill uses the GitHub CLI to fetch the PR title, body, and full diff for analysis.

Does it analyze the whole file or just the diff?

The agents read the full content of any modified files to understand the context of the changes, rather than just looking at isolated lines in a diff.

Automated Code Review

Name: Automated Code Review
Author: jjeremycai

byjjeremycai

0•

コラボレーションツール

Performs comprehensive code audits using specialized subagents to identify logic bugs, security vulnerabilities, and pattern violations.

The Automated Code Review skill streamlines the development workflow by deploying three parallel subagents focused on distinct areas: logic/edge cases, security/adversarial risks, and codebase/architectural fit. It automatically analyzes uncommitted local changes, the most recent commit, or specific GitHub Pull Requests to provide a severity-ranked summary of issues. By surfacing unstated assumptions and providing an integrated path to auto-fix identified problems, it ensures high-quality code and reduces the burden on human reviewers.

主な機能

01Support for local uncommitted changes, recent commits, and GitHub PRs

020 GitHub stars

03Triple-agent analysis: Bug Hunter, Security, and Codebase Fit specialists

04Assumption surfacing to identify potential future logic failures

05Severity-ranked reporting (Critical, Warning, and Suggestion levels)

06Integrated /fix workflow for automated resolution of identified issues

ユースケース

01Identifying unstated assumptions in legacy code or complex refactors

02Pre-commit sanity checks to catch logic errors and security leaks early

03Automated Pull Request auditing to ensure compliance with project patterns

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jjeremycai/claudekit review

For use in Claude.ai and ChatGPT

Download Skill