What problem does it solve? New agent skills often ship untested, with ambiguous instructions, missing tool fallbacks, and unclear triggers that only surface when a real agent runs them. This Skill enforces a structured trial-run process so every skill is validated by a fresh agent before merging. ## Core Features & Use Cases - Five-Phase Workflow: Ideate features and tools, build SKILL.md plus reference files, run a trial in a separate agent thread, review the trial thread for failures, then iterate and open a PR. - Structured Trial Review: Classifies trial issues into categories like tool failures, ambiguous instructions, missing fallbacks, monorepo issues, and output quality, then maps each to a fix target. - Quality Gates: Enforces a build checklist covering frontmatter validity, description triggers, cross-reference integrity, line limits under 500, and skill-lint execution when available. - Use Case: When you say "build a skill for running security audits", the Skill brainstorms the feature set with you, authors the SKILL.md, has you trial it in a clean agent thread, reads that thread for friction points, and iterates until the skill passes all quality checks. ## Quick Start Ask the agent to build a new skill for your chosen task using the trial-run workflow and approve the proposed scope when prompted.
