writing-skills

Test-driven development for skills means writing baseline tests first to observe how agents fail without the skill, then drafting documentation and iterating until tests pass. This ensures your skill documentation translates to actual agent compliance, not just theory.

Pressure scenarios combine time constraints, sunk costs, authority pressure, and other real-world factors to test whether agents follow your skill rules under stress. They expose rationalization loopholes that normal testing misses.

Structure your skill with strict YAML frontmatter containing name and description within length limits, then iterate through RED-GREEN-REFACTOR cycles: write failing tests, make them pass, refactor. This ensures ASO-aligned metadata and functional requirements alignment.

Documenting exact rationalizations agents produce when they deviate reveals the specific loopholes your skill must close. This prevents agents from inventing workarounds and strengthens your rules against real pressure scenarios.

Yes. The skill guides you through creating pressure scenarios with subagents to establish baseline failures, then iterating documentation until subagents pass all tests under realistic constraints.

AI skill documentation must pass compliance tests under pressure—agents must actually follow the rules, not just understand them. Regular documentation assumes human reading; skill documentation requires observable, measurable agent behavior change.