testing-skills-with-subagents

TDD-driven testing validates skill behavior through Red-Green-Refactor cycles with Serena metrics. Run baseline scenarios to identify failures, implement fixes, then re-run to achieve compliance_score > 0.85, ensuring bulletproof performance in high-pressure deployment environments.

Serena metrics quantify compliance across scenarios involving time pressure, sunk cost, authority, exhaustion, and social pressures that may induce rationalizations. They track baseline failures, compliance scores, and loopholes to verify process discipline and continuous improvement through meta-testing.

Log baseline failures, calculate compliance_score across pressure scenarios, track loopholes, and run meta-tests to validate prompt clarity. Iterate through Red-Green-Refactor cycles until the final score exceeds 0.85, ensuring consistent behavior across subagent deployment conditions.

The framework tests skills against five pressure vectors: time pressure, sunk cost fallacy, authority influence, exhaustion, and social pressure. These scenarios simulate real deployment conditions where rationalization risks are highest, validating that documented processes hold under stress.

The Skill enforces TDD discipline through structured Red-Green-Refactor cycles, making it accessible to TDD learners. Start with baseline scenarios, implement the skill to address failures, then validate compliance—the cycle itself teaches TDD fundamentals while building pressure-resistant skills.

Meta-testing validates that prompts and process documentation are clear and unambiguous. It catches rationalizations or loopholes in skill logic, ensuring the final compliance score reflects genuine robustness rather than incomplete coverage.