概要
Skill Testing with Subagents is a specialized framework that brings the rigors of Test-Driven Development (TDD) to the creation of Claude skills. By following a strict Red-Green-Refactor cycle, developers can verify that their custom instructions and process documentation actually change agent behavior in high-pressure, real-world scenarios. The skill focuses on identifying 'rationalization loopholes'—the specific excuses AI agents use to bypass rules—and provides structured methods like pressure testing and rationalization tables to plug those holes, ensuring mission-critical instructions are followed without exception.
