Table of Contents

• Overview
• Tools
• What It Is
• Quick Start
• Basic Skill Audit
• Skill Analysis
• Generate Improvements
• Typical Workflow
• Common Tasks
• Quality Assessment
• Performance Analysis
• Standards Compliance
• Improvements and Optimization
• Evaluation Framework
• Quality Metrics Overview
• Scoring System
• Priority Levels
• Detailed Resources
• Shared Modules (Cross-Skill Patterns)
• Skill-Specific Modules
• Tools and Automation

Skills Evaluation and Improvement

Overview

Analyze and improve Claude skills. Tools audit skills against quality standards, measure token usage, and generate improvement recommendations.

Tools

The evaluation process uses skills-auditor for structural analysis and improvement-suggester for prioritized fixes. compliance-checker validates standards, tool-performance-analyzer measures tool patterns, and token-usage-tracker monitors context efficiency.

What It Is

Evaluates and improves existing skills by running quality assessments, performance analysis, and generating improvement plans.

Quick Start

Basic Skill Audit

# Run detailed audit of all skills
python scripts/skills_eval/skills_auditor.py --scan-all --format markdown

# Audit specific skill
python scripts/skills_eval/skills_auditor.py --skill-path path/to/skill/SKILL.md

# Or use Makefile:
make audit-skill PATH=path/to/skill/SKILL.md
make audit-all

Skill Analysis

# Deep analysis of single skill
python scripts/skill_analyzer.py --path path/to/skill/SKILL.md --verbose

# Check token usage
python scripts/token_estimator.py --file path/to/skill/SKILL.md

# Or use Makefile:
make analyze-skill PATH=path/to/skill/SKILL.md
make estimate-tokens PATH=path/to/skill/SKILL.md

Generate Improvements

# Get prioritized improvement suggestions
python scripts/skills_eval/improvement_suggester.py --skill-path path/to/skill/SKILL.md --priority high

# Check standards compliance
python scripts/skills_eval/compliance_checker.py --skill-path path/to/skill/SKILL.md --standard all

# Or use Makefile:
make improve-skill PATH=path/to/skill/SKILL.md
make check-compliance PATH=path/to/skill/SKILL.md

Typical Workflow

1. Discovery: Run make audit-all to find and audit all skills
2. Analysis: Use make audit-skill PATH=... for specific skills
3. Deep Dive: Run make analyze-skill PATH=... for complexity analysis
4. Improvements: Generate plan with make improve-skill PATH=...
5. Compliance: Verify standards with make check-compliance PATH=...
6. Optimization: Check tokens with make estimate-tokens PATH=...

Common Tasks

Quality Assessment

# Detailed evaluation with scoring
./scripts/skills-auditor --scan-all --format table --priority high

# Detailed analysis of specific skill
./scripts/improvement-suggester --skill-path path/to/skill/SKILL.md --priority all --format markdown

Performance Analysis

# Token usage and efficiency
./scripts/token-usage-tracker --skill-path path/to/skill/SKILL.md --context-analysis

# Advanced tool performance metrics
./scripts/tool-performance-analyzer --skill-path path/to/skill/SKILL.md --metrics all

Standards Compliance

# Validate against Claude Skills standards
./scripts/compliance-checker --skill-path path/to/skill/SKILL.md --standard all --format summary

# Auto-fix common issues
./scripts/compliance-checker --skill-path path/to/skill/SKILL.md --auto-fix --severity high

Improvements and Optimization

# Generate prioritized improvement plan
./scripts/improvement-suggester --skill-path path/to/skill/SKILL.md --priority critical,high

# Benchmark performance
./scripts/token-usage-tracker --skill-path path/to/skill/SKILL.md --benchmark optimization-targets

Evaluation Framework

Quality Metrics Overview

The framework evaluates skills on weighted dimensions: structure compliance and content quality (20 points each), token efficiency and activation reliability (15 points each), and remaining points for tool integration, trigger isolation, and enforcement language.

Scoring System

Scores categorize quality: 91-100 (Excellent), 76-90 (Good), 51-75 (Basic), 26-50 (Below Standards), and 0-25 (Critical Issues).

Priority Levels

Improvements are prioritized to address the most critical issues first. Critical priority is assigned to security vulnerabilities, broken functionality, or missing required metadata. High priority concerns poor structure or incomplete documentation, while medium priority identifies missing best practices and optimization opportunities. Minor enhancements, such as formatting issues or improved examples, are categorized as low priority.

Detailed Resources

For detailed implementation details and advanced techniques:

Shared Modules: Cross-Skill Patterns

• Anti-Rationalization Patterns: See anti-rationalization.md for red flags table and bypass patterns
• Enforcement Language: See enforcement-language.md for tiered intensity templates
• Trigger Patterns: See trigger-patterns.md for description field structure and CSO

Skill-Specific Modules

• Trigger Isolation Analysis: See modules/trigger-isolation-analysis.md for evaluating frontmatter compliance
• Skill Authoring Best Practices: See modules/skill-authoring-best-practices.md for official Claude guidance
• Authoring Checklist: See modules/authoring-checklist.md for quick-reference validation checklist
• Implementation Guide: See modules/evaluation-workflows.md for detailed workflows
• Quality Metrics: See modules/quality-metrics.md for scoring criteria and evaluation levels
• Advanced Tool Use Analysis: See modules/advanced-tool-use-analysis.md for specialized evaluation techniques
• Evaluation Framework: See modules/evaluation-framework.md for detailed scoring and quality gates
• Integration Patterns: See modules/integration.md for workflow integration with other skills
• Troubleshooting: See modules/troubleshooting.md for common issues and solutions
• Pressure Testing: See modules/pressure-testing.md for adversarial validation methodology

Tools and Automation

• Tools: Executable analysis utilities in scripts/ directory
• Automation: Setup and validation scripts in scripts/automation/