write-skills

To apply TDD to skill creation, draft a minimal SKILL.md, write baseline RED tests without the skill, then implement and re-test until all scenarios pass. This enforces frontmatter validation and closes loopholes before deployment.

The RED-GREEN-REFACTOR workflow for writing agent skills involves writing baseline tests first (RED), implementing the skill to pass tests (GREEN), and refactoring to close loopholes. It prevents gaps and inconsistent discovery in skill documentation.

Agent skill frontmatter validation fails when the name and description exceed length limits or lack required schema fields. Applying a disciplined TDD workflow checks frontmatter validation early, preventing deployment errors from inconsistent documentation.

Yes, you can use TDD to verify skill documentation with subagents. The workflow provides guidance for testing skills with subagents and meta-testing, ensuring the documentation behaves correctly and consistently before final deployment.

The best way to write reliable agent skills without rationalizations is enforcing a test-driven documentation process. Standardized frontmatter schemas and evaluation patterns prevent gaps, ensuring reusable and error-free skill deployment.

Yes, there are length limits for agent skill frontmatter name and description fields. Enforcing these frontmatter length limits during the RED-GREEN-REFACTOR workflow ensures standardized evaluation patterns and prevents discovery inconsistencies.