What problem does it solve? Agents often fail at tasks without proper guidance, and poorly written skills either never trigger or get ignored. This Skill provides a rigorous methodology for authoring, testing, and deploying skills so they reliably change agent behavior. ## Core Features & Use Cases - Skill Anatomy Standards: Defines the required SKILL.md frontmatter (name, description, keywords, paths) and the progressive disclosure structure with optional scripts, references, and assets directories. - TDD-Based Creation Process: Enforces a RED-GREEN-REFACTOR cycle where you must observe an agent failing without the skill before writing it, then close rationalization loopholes. - Claude Search Optimization: Rules for writing trigger-focused descriptions ("Use when...") that avoid workflow summaries, plus keyword coverage and naming conventions. - Use Case: When you notice an agent repeatedly mishandling a task, use this Skill to baseline the failure, author a minimal skill addressing it, pressure-test it, and deploy it via compile-harness.sh. ## Quick Start Ask the agent to create a new skill for a recurring failure pattern, following the TDD baseline-first process defined in this skill.
