What problem does it solve?
Manually reviewing hundreds of GitHub PR review comments to identify recurring code issues is time-consuming and prone to oversight, especially for teams handling dozens of PRs weekly. This Skill automates the extraction of consistent patterns from review feedback to help teams improve coding standards and reduce repeat issues.
Core Features & Use Cases
- Bounded 3-Phase Processing: Uses a collect/fan-out/aggregate architecture to process large volumes of PR comments without overwhelming context, with configurable caps on PR count, comment length, and parallel subagent tasks.
- Cross-PR Pattern Deduplication: Identifies patterns that appear across multiple PRs to flag high-priority recurring issues, and cross-references existing team pattern files to avoid duplicates.
- Learning System Integration: Automatically feeds extracted patterns into the team's learning system for review, with optional approval workflows to update coding standard rule files.
- Use Case: For an engineering team running 20+ PRs a week, use this Skill to automatically surface common issues like missing validator wiring or ETag propagation bugs across all recent PRs in minutes instead of hours.
Quick Start
Use the pr-pattern-extract skill to analyze the last 5 merged PRs in your current repository and surface recurring code review patterns.