What problem does it solve?
Ensures AI skill authors produce discoverable, concise, and testable SKILL.md documentation that agents reliably follow under real-world pressures and temptations to bypass rules.
Core Features & Use Cases
- Test-Driven Skill Authoring: Maps RED-GREEN-REFACTOR to skill creation so authors write failing baseline scenarios, create minimal skills, and iteratively close loopholes.
- Pressure Testing & Verification: Provides templates and scenarios to simulate time, sunk-cost, and authority pressures so skills enforce discipline when compliance is hardest.
- Discovery & CSO Guidance: Prescribes frontmatter, description design, keyword coverage, and token-efficient authoring to help agents find and load the right skill when needed.
Quick Start
Run a RED-GREEN-REFACTOR check: execute a realistic pressure scenario without the skill, document exact failures, write or update SKILL.md to address those failures, and re-run the scenario until the agent complies.