testing-skills-with-subagents

TDD applies to process documentation by using a RED-GREEN-REFACTOR cycle to test skills. This identifies baseline failures, writes minimal skills to fix them, and refactors to close loopholes against agent rationalizations.

The best way to ensure agent compliance under pressure is through structured pressure testing. This simulates realistic scenarios where agents are tempted to break rules, allowing you to close loopholes and strengthen the skill's effectiveness.

To start the RED-GREEN-REFACTOR cycle for skill testing, run the testing framework. It begins with baseline testing to identify failures without the skill, then writes minimal skills to address those specific failures.

You need baseline testing before writing an agent skill to identify failures that occur without the skill. This establishes what specific problems need preventing and guides the creation of minimal, effective skills to address them.

To close loopholes in agent skills when rationalization occurs, use the refactoring phase of the testing cycle. This strengthens the skill by analyzing rationalizations under pressure and updating the documentation to prevent future bypasses.

Yes, you can test process documentation effectiveness before deployment by applying this skill testing framework. It subjects your skills to realistic pressure scenarios to verify compliance and close loopholes prior to release.