skill-writing-tdd

Test-driven development for skill authoring requires observing baseline failures without the skill, writing a minimal skill to pass tests, then refactoring to close loopholes with explicit counters. This RED-GREEN-REFACTOR cycle ensures skills are proven compliant before deployment.

The RED-GREEN-REFACTOR discipline is a testing lifecycle for skills where RED observes baseline failures, GREEN writes a minimal skill to pass, and REFACTOR closes loopholes using anti-rationalization tables and explicit counters. It guarantees deployment safety and reusability.

Validating skill quality and diagnosing agent behavior requires iterative testing with subagents and deployment checklists. By observing baseline failures and applying four-state reporting with HARD-GATE markers, you verify compliance and prevent rationalization before deployment.

Anti-rationalization tables close loopholes in skill documentation by providing explicit counters and HARD-GATE markers during the refactoring phase. This prevents agents from bypassing rules and enhances the reliability of the tested skill.

The best way to prevent rationalization when writing skills is enforcing a disciplined lifecycle using anti-rationalization tables, red flags, and four-state reporting. Refactoring documentation with explicit counters closes loopholes and ensures bulletproof guides.

Test-driven development for skill documentation should not be used when you cannot observe baseline failures or run iterative tests with subagents. Without proving a minimal skill passes compliance checks, applying the RED-GREEN-REFACTOR cycle provides no validation value.