Evaluation-Driven Skill Development

소개

This skill provides a structured methodology for building agentic AI skills using an Evaluation-Driven Development (EDD) approach. It walks developers through creating evaluation scenarios before writing documentation, establishing performance baselines, conducting multi-model testing via the Claude Code CLI, and performing final reviews to ensure compliance with best practices. By focusing on measurable outcomes rather than instruction content, it ensures that new skills effectively solve real-world problems and work reliably across different Claude models.

주요 기능

• Baseline performance measurement and gap analysis
• Step-by-step EDD workflow for AI skills
• Integrated skill review and optimization tools
• Automated multi-model evaluation integration
• Automated generation of user validation guides
• 0 GitHub stars

사용 사례

• Establishing a testing framework for new agentic skills
• Refining existing skills based on observed failure patterns
• Validating skill performance across Sonnet, Opus, and Haiku models