pr-reviewer

Perform adversarial multi-pass PR reviews and report findings in findings.md.

2|Updated Mar 11, 2026
One-click install
npx skills add https://github.com/mdelapenya/coding-skills --skill pr-reviewer-mdelapenya
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pr-reviewer
Source: https://github.com/mdelapenya/coding-skills/tree/main/skills/pr-reviewer
Command: npx skills add https://github.com/mdelapenya/coding-skills --skill pr-reviewer-mdelapenya

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This PR reviewer helps teams find real problems by performing an adversarial, multi-pass review and forcing self-challenge to reduce false positives.

Core Features & Use Cases

  • Adversarial first pass that identifies potential issues and invites a second self-challenge pass.
  • Looped convergence to continue reviewing until findings converge, with optional rounds and a single-round mode.
  • Flexible review contexts: run inside the matching repo or from a barebones directory, cloning as needed and keeping state in findings.md.

Quick Start

Invoke the PR reviewer on the current branch or a specified PR to begin a multi-pass adversarial review and save findings to findings.md.

Frequently Asked Questions about pr-reviewer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an adversarial code review on a pull request to find real issues?

An adversarial code review identifies potential issues in a pull request and forces a self-challenge pass to reduce false positives, surfacing real problems by converging findings across multiple review rounds.

What is the best way to automate multi-pass PR reviews until findings converge?

Multi-pass PR reviews use a looped convergence mechanism to continue reviewing until findings stabilize. You can control this process by setting the number of rounds or stopping it after a single pass.

Can I review a pull request from a bare directory without cloning the repository first?

Yes, pull request reviews can run from a bare directory. The process automatically clones the PR head into a temporary directory when needed, so you do not need to manually clone the repository beforehand.

How does an adversarial PR review track state across multiple review rounds?

Adversarial PR reviews track state across rounds by writing structured findings to a findings.md file. This ensures review progress and identified issues persist throughout the convergence loop.

How do I stop a looped pull request review after a single pass?

You can stop a looped pull request review after a single pass by using the no-loop control flag. Alternatively, you can specify an exact number of rounds to limit the convergence loop duration.

What metadata does a rigorous pull request review gather before analyzing diffs?

A rigorous pull request review gathers PR metadata, commits, diffs, linked issues, and touched files before analysis. This comprehensive context supports the adversarial identification of real issues in the changes.