customaize-agent:test-skill

You validate skill documentation by running automated RED baseline tests to capture agent failures, implementing GREEN corrective content, and refining rules to confirm resilience under simulated pressure.

The RED-GREEN-REFACTOR cycle for skill testing is a validation process where RED captures baseline agent failures, GREEN implements corrective content, and REFACTOR tightens rules to close loopholes and maintain compliance.

Yes, TDD-style validation requires a SKILL.md file containing YAML frontmatter with name and description fields, and optionally supports scripts, references, and assets for extended testing scenarios.

You run baseline RED tests by executing validation scenarios against the documentation without the skill applied, which captures agent failures and establishes a failure baseline for corrective implementation.

The best way to harden skill content is applying meta-testing through a REFACTOR phase that tightens rules, closes loopholes, and maintains compliance to verify clarity and resilience under simulated pressure.

Skill documentation fails validation because content resists rationalization poorly without baseline testing, allowing agents to bypass intended rules when simulated pressure exposes unstated assumptions and loopholes.