dyad:multi-pr-review

Coordinate three Claude agents to review PR diffs via consensus voting.

420|65|Updated Jan 24, 2026
One-click install
npx skills add https://github.com/CoWork-OS/CoWork-OS --skill dyad-multi-pr-review-cowork-os
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: dyad:multi-pr-review
Source: https://github.com/CoWork-OS/CoWork-OS/tree/main/resources/skills/multi-pr-review
Command: npx skills add https://github.com/CoWork-OS/CoWork-OS --skill dyad-multi-pr-review-cowork-os

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires anthropic, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Multi-agent PR review orchestrates three independent Claude sub-agents to review PR diffs. Each agent receives files in different randomized order to reduce ordering bias. One agent focuses specifically on code health and maintainability. Issues are classified as high/medium/low severity (sloppy code that hurts maintainability is MEDIUM). Results are aggregated using consensus voting - only issues identified by 2+ agents where at least one rated it medium or higher severity are reported. Automatically deduplicates against existing PR comments. Always posts a summary (even if no new issues), with low priority issues mentioned in a collapsible section.

Core Features & Use Cases

  • Spawns three independent sub-agents, with randomized file orderings to diversify perspectives
  • Uses consensus voting to aggregate findings and report only agreed issues
  • Deduplicates against existing PR comments and posts a summary plus inline feedback

Quick Start

Start the orchestrator with your PR number and diff to begin tri-agent review.

Frequently Asked Questions about dyad:multi-pr-review

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

FAQPage Schema
How does multi-agent consensus voting work for PR review?

Multi-agent consensus voting for PR review spawns three independent sub-agents to analyze diffs and only reports issues identified by at least two agents, ensuring agreed findings are surfaced while reducing individual bias.

How do I automate deduplication of code review comments on a PR diff?

You can automate deduplication of code review comments by running a multi-agent review that cross-references newly identified diff issues against existing PR comments, posting only unique inline feedback and a summary.

Do I need an Anthropic API key to run automated multi-agent PR review?

Yes, you need an Anthropic API key to run automated multi-agent PR review, as the sub-agents rely on Anthropic models to analyze diffs, classify issue severity, and coordinate consensus voting.

Can I configure the number of agents and consensus threshold for diff analysis?

Yes, you can configure the agent count and consensus rules for diff analysis using parameters like NUM_AGENTS and CONSENSUS_THRESHOLD, alongside MIN_SEVERITY and THINKING_BUDGET to control decision depth.

What is the best way to reduce ordering bias in automated code review?

The best way to reduce ordering bias in automated code review is to provide randomized file orderings to multiple independent review agents, diversifying their perspectives before aggregating results through consensus.

How are low severity code health issues handled in automated PR feedback?

Low severity code health issues are placed in a collapsible section within the PR summary, while sloppy code affecting maintainability is classified as MEDIUM severity to ensure it meets the consensus reporting threshold.