How does the skill know which feedback is worth saving?

It uses a multi-step logic to filter for 'Resolved' threads or comments that were followed by a corrective commit, ensuring only successfully addressed feedback becomes a rule.

Can I choose which rules are added to CLAUDE.md?

Yes, the skill presents a list of suggested rules and categories for your approval before any modifications are made to your files.

Does it work with private repositories?

Yes, it works with any repository accessible via your local GitHub CLI configuration.

Does this skill require specific permissions?

It requires the GitHub CLI (gh) to be installed and authenticated with read access to your repository's Pull Requests and comments.

How does this differ from the standard /lessons command?

While /lessons captures self-identified mistakes during a session, /review-learn captures external feedback from teammates, combining personal growth with team-wide standards.

Review Feedback Learner

Name: Review Feedback Learner
Author: xtone

byxtone

•

Collaboration Tools

Extracts actionable insights from GitHub PR reviews and transforms them into persistent project rules in CLAUDE.md.

Review Feedback Learner is a specialized capability designed to bridge the gap between human code reviews and AI-assisted development. It automates the process of codifying 'implicit knowledge' by scanning resolved Pull Request comments and converting them into structured rules within your project's CLAUDE.md file. By analyzing reviewer feedback on naming conventions, architecture, and error handling, the skill ensures that once a mistake is corrected in a PR, Claude learns the standard and prevents it from happening again. This creates a self-improving development environment where team standards are enforced by the AI in real-time.

Key Features

01Intelligent filtering that distinguishes between bot noise and high-value human insights.

02Direct attribution features that link rules back to original PRs for historical context.

03Smart duplication checking to ensure CLAUDE.md remains concise and relevant.

04Automated PR scanning using GitHub GraphQL API to identify resolved and approved feedback.

053 GitHub stars

06Automatic categorization of feedback into themes like Architecture, Security, and Naming Conventions.

Use Cases

01Reducing 'reviewer fatigue' by ensuring common nits and recurring patterns are handled by the AI before the PR is even opened.

02Codifying team-specific coding standards that are often caught during manual PR reviews.

03Onboarding new developers by providing an AI that already knows the project's 'unwritten' rules.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xtone/ai_development_tools review-feedback-learner

For use in Claude.ai and ChatGPT

Download Skill