What problem does it solve? Untested skill documentation often fails in practice because agents rationalize around rules under time, sunk-cost, or authority pressure, and authors cannot know whether a skill teaches the right behavior without observing baseline failures first. ## Core Features & Use Cases - TDD-based skill authoring: Applies the RED-GREEN-REFACTOR cycle to documentation by running baseline pressure scenarios with subagents, writing the skill to address observed failures, and closing loopholes iteratively. - Discovery optimization guidance: Provides rules for frontmatter descriptions, keyword coverage, naming conventions, and token efficiency so agents can find and load the right skill. - Testing methodology references: Includes detailed guidance on pressure scenario design, rationalization tables, persuasion principles, and Anthropic's official authoring best practices. - Use Case: Before deploying a new discipline-enforcing skill like a TDD rule, run combined-pressure scenarios (time + sunk cost + exhaustion) against subagents without the skill, capture their verbatim rationalizations, then write and refactor the skill until agents comply under maximum pressure. ## Quick Start Ask the agent to create a new skill for a recurring technique, first running baseline pressure scenarios with subagents and then writing the SKILL.md following the RED-GREEN-REFACTOR checklist.
