writing-skills

Test-driven skill authoring applies RED-GREEN-REFACTOR cycles to documentation by capturing baseline agent failures, writing content to address them, then refining to close loopholes. This ensures validated, repeatable compliance before deployment.

RED-GREEN-REFACTOR for skill documentation involves drafting YAML frontmatter, writing RED scenarios to capture baseline failures, implementing GREEN content to fix them, and refactoring to close new loopholes and prevent rationalizations.

Agent skills fail under pressure when baseline testing patterns are skipped, leaving loopholes for rationalizations. Observing agent behavior without a skill first identifies concrete failure modes to systematically close during iterative refinement.

Validate Claude Skills before deployment by enforcing iterative RED-GREEN-REFACTOR test cycles. Skills reach deployment readiness only after achieving repeatable green outcomes that confirm compliance under pressure and prevent rationalizations.

Yes, TDD methodology applies to authoring, editing, and pre-deployment validation of skills. You update counters and red flags in existing structured sections, then re-run tests to verify compliance and close newly identified loopholes.

TDD for skill authoring requires drafting structured YAML frontmatter and documenting rationalizations verbatim. It is unsuitable if you cannot observe baseline agent failures or lack resources for iterative refinement to achieve repeatable green outcomes.