소개
This skill introduces a rigorous framework for authoring Claude Code skills by treating documentation like production code. By adapting the Red-Green-Refactor cycle, it ensures that every skill is verified against real-world AI behavior. Users first watch a subagent fail a task without the skill, write the documentation to address specific failure points, and then iterate until the AI complies without rationalization. This methodology results in highly effective, token-efficient, and reliable domain-specific guidance for future Claude instances.
