testing-skills-with-subagents

To test skills against rationalization, run baseline scenarios without the skill to capture verbatim failures, then apply a RED-GREEN-REFACTOR cycle to iteratively refine the documentation and close loopholes.

The RED-GREEN-REFACTOR cycle for skill documentation involves identifying baseline failures, writing targeted documentation to address them, and iteratively refining the text to close loopholes before deployment.

You capture baseline failures by executing structured testing scenarios without the skill active, recording the exact rationalization failures verbatim to surface documentation gaps before iterating on fixes.

Yes, you can apply TDD principles to quality assurance workflows by running a baseline without the skill, documenting failures, and using repeatable verification across multiple scenarios to validate iterative refinements.

You should run meta-testing on skills before deployment to verify they work under pressure, tracking rationalization failures and applying a structured RED-GREEN-REFACTOR loop to validate the documentation.

The limitation of this structured testing workflow is that it requires repeatable verification across multiple scenarios and demands rigorous rationalization tracking to ensure documentation holds up under pressure.