writing-skills

Test-driven development for AI skill creation enforces a RED-GREEN-REFACTOR cycle to author and refine documentation, requiring baseline failure analysis before writing minimal skill logic and iteratively refactoring to close loopholes.

Pressure testing for agent behavior involves designing scenarios that test skills under realistic constraints like time pressure, sunk cost, and authority to ensure the agent adheres to rules and resists manipulation.

To stop agent rationalization, you must identify common excuses and explicitly address them in the documentation, using iterative refactoring to close loopholes against rationalization during skill development.

The best way to document robust AI skills is to follow a rigorous test-driven development framework, ensuring documentation remains compliant by validating it against baseline failures and pressure scenarios.

You should use a TDD methodology for refining AI skills before deploying critical coding standards, allowing you to simulate developers trying to skip standards under deadline pressure and verify the skill holds up.

The RED-GREEN-REFACTOR cycle requires no external dependencies, relying solely on internal scripts and references to enforce test-driven development for all skill modifications and documentation compliance.