How does the PR Review skill minimize false positives?

The skill uses a multi-step validation process where any issue found by a primary agent must be independently verified by a validation subagent and meet a confidence score of at least 80/100 before being reported.

Can it follow my project's specific coding standards?

Yes, the skill includes a Guidelines Discovery agent that automatically finds and reads CLAUDE.md files and other local style guides to ensure feedback aligns with your project's unique conventions.

Does this skill require manual GitHub CLI configuration?

It relies on the GitHub CLI (`gh`) being authenticated on your system, as it uses automated scripts to fetch PR details and post comments on your behalf.

Will it spam my PR with dozens of individual notifications?

No, the skill uses a batched review workflow. It accumulates all validated feedback into a single JSON structure and posts it as one cohesive GitHub review.

PR Review Agent

Name: PR Review Agent
Author: enitrat

byenitrat

•

Collaboration Tools

Automates thorough, high-signal pull request reviews using parallel specialized agents and rigorous validation.

The PR Review skill orchestrates a sophisticated multi-agent workflow to provide professional-grade code reviews. By deploying specialized agents for bug hunting, guideline compliance, and error-handling audits in parallel, it ensures comprehensive coverage of every change. To maintain high signal and low noise, the skill includes a dedicated validation phase where findings are double-checked for accuracy. It integrates seamlessly with the GitHub CLI to perform pre-review checks and post-batched, structured feedback, allowing teams to maintain high code quality standards while accelerating the merge process.

Key Features

01Parallel multi-agent architecture using Claude 3 Opus and Sonnet

02Batched GitHub comment posting via GitHub CLI to prevent notification spam

03Pre-review validation to skip draft, closed, or trivial pull requests

04Automated discovery of project-specific guidelines and CLAUDE.md rules

05Confidence-based issue validation to minimize false positives

064 GitHub stars

Use Cases

01Enforcing architectural consistency and pattern reuse across large codebases

02Automating the initial review pass to catch objective bugs and style violations

03Performing deep-dive security and error-handling audits on critical PRs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add enitrat/skill-issue pr-review

For use in Claude.ai and ChatGPT

Download Skill