testing-skills-with-subagents

Baseline testing observes system failures and rationalizations by running scenarios without the skill applied. This establishes a control group to identify exactly which failure modes the skill needs to prevent.

Apply the RED-GREEN-REFACTOR cycle by baseline testing failures without the skill, writing skill instructions to fix those failures, pressure testing for loopholes, and refactoring the documentation to close them.

The best way to prevent rationalization is through pressure testing. By creating realistic test scenarios that attempt to bypass the rules, you identify loopholes in the documentation and refactor it to close them.

Yes, you need an understanding of Test-Driven Development principles and the ability to create realistic test scenarios to effectively apply this cycle to process documentation and skill verification.

You should test skills under pressure before deploying any new skill. This ensures the process documentation will work effectively and maintain compliance in real-world scenarios by closing rationalization loopholes beforehand.