Agent Skill Quality Reviewer

0•

Audits and improves AI agent skills to ensure they meet high standards for clarity, correctness, and effectiveness.

The bd-skill-reviewing skill provides a comprehensive framework for evaluating and optimizing Claude Code agent skills. It guides users through a systematic review process focusing on four key dimensions: correctness, clarity, effectiveness, and maintainability. By utilizing standardized checklists and identifying common anti-patterns, this skill helps developers troubleshoot underperforming agents, provide constructive feedback on pull requests, and ensure all skills adhere to organizational quality standards and technical specifications.

주요 기능

01Standardized feedback templates for consistent and actionable PR reviews
02Improvement guidance with actionable before-and-after implementation patterns
03Detailed quality checklists for structure and frontmatter validation
040 GitHub stars
05Systematic review framework across four critical quality dimensions
06Anti-pattern detection to identify and fix common skill weaknesses

사용 사례

01Troubleshooting and optimizing underperforming or ambiguous agent instructions
02Performing quality assurance on new AI agent skills before deployment
03Conducting structured peer reviews on skill-related pull requests
name description
bd-skill-reviewing
Framework for reviewing and improving agent skills for quality, effectiveness, and compliance with the agent-skills standard. Use when auditing existing skills, providing feedback on skill PRs, improving skill descriptions, or troubleshooting underperforming skills.

Skill Review Guide

This skill provides a systematic framework for evaluating and improving agent skills.

Review Criteria

Evaluate skills against four dimensions:

  1. Correctness: Follows specification requirements
  2. Clarity: Instructions are unambiguous and actionable
  3. Effectiveness: Triggers appropriately and guides well
  4. Maintainability: Well-organized and easy to update

Review Workflow

  1. Validate Structure: Check directory and file organization
  2. Evaluate Frontmatter: Verify name and description compliance
  3. Assess Content: Review clarity, completeness, progressive disclosure
  4. Check References: Verify organization and linking
  5. Test Effectiveness: Confirm triggers and guidance work

For detailed checklist with specific checks per category, see references/quality-checklist.md.

Common Issues

For anti-patterns to identify during review (with detection methods and fixes), see references/anti-patterns.md.

Improvement Guidance

For patterns to fix common issues (with before/after examples), see references/improvement-patterns.md.

Feedback Format

For review feedback templates and severity guidelines, see references/feedback-template.md.

Anti-Patterns in Reviews

  • Rubber-stamping: Approving without thorough review
  • Style-only focus: Missing structural or effectiveness issues
  • No actionable feedback: Identifying problems without suggesting fixes
  • Inconsistent severity: Misclassifying critical issues as minor