Where are the review results stored?

Reviews are stored locally in the ~/claude-loop-pr-copilot/ directory, organized by repository and PR number, including individual logs and the final unified report.

How does the skill select which PR to review?

It uses the GitHub Search API to find open PRs where you are a requested reviewer, filtering out those you have already approved or those currently being processed.

Does this skill require both Claude Code and Copilot CLI?

Yes, this skill is designed as a dual-reviewer system that orchestrates both Claude Code and the Copilot CLI to provide a more balanced and thorough code analysis.

Can this skill handle private repositories?

Yes, as long as your local GitHub CLI (gh) is authenticated and has the necessary permissions to access and clone the repositories.

How does it handle conflicting feedback between the two AIs?

The skill includes a synthesis step where the main agent compares the findings, highlighting points of agreement as high-confidence issues and noting differences for the human developer to evaluate.

PR Copilot Dual Review

Name: PR Copilot Dual Review
Author: yuki777

byyuki777

0•

协作工具

Automates GitHub Pull Request reviews by synthesizing insights from both Claude Code and Copilot CLI for comprehensive code analysis.

This skill streamlines the code review process by delegating PR analysis to two powerful AI engines—Claude Code and Copilot CLI. It automatically scans your GitHub account for pending review requests, identifies eligible open PRs, and performs parallel reviews using isolated working directories to ensure accuracy and speed. By merging the findings from both tools into a single, structured report, it provides a high-confidence summary of critical bugs, security risks, and style improvements, allowing developers to focus on high-level architectural decisions rather than routine syntax checks.

主要功能

01Synthesis of results into a unified markdown report with consensus tracking

02Automated discovery of GitHub PRs where the user is a requested reviewer

030 GitHub stars

04Parallel review execution with isolated clone management

05Dual-engine analysis using Claude Code and Copilot CLI

06State management via status.json to track pending, running, and failed reviews

使用场景

01Generating objective 'second opinions' on complex PRs from multiple AI models

02Automating the first pass of code reviews to identify critical issues before human inspection

03Managing a high volume of review requests across multiple repositories efficiently

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yuki777/pr-copilot review

For use in Claude.ai and ChatGPT