What problem does it solve? Teams building their own AI skill collections often duplicate existing skills, write descriptions that never trigger, ship instructions nobody follows, and let packaged archives drift out of sync with source. This Skill provides the authoring discipline that prevents those failure modes. ## Core Features & Use Cases - Search-before-build workflow: Checks local skills, marketplaces, GitHub, and the web (in cost order) with a security vetting pass on any external skill before adoption. - Quality bar and compliance testing: Defines what makes a skill done (triggering description, stated boundaries, provenance) and measures whether agents actually obey a skill using supportive, neutral, and competing prompt levels. - Packaging and canonical contracts: Validates frontmatter, packages skills into deterministic dist/*.skill archives via package_skills.py, and hosts shared contracts for language routing, the seven-stage development loop, Context7 docs lookup, and surgical code changes. - Use Case: Before adding a new skill to the repo, run the search sequence to confirm nothing similar exists, write it to the quality bar, then run python scripts/package_skills.py <name> so CI's drift check passes. ## Quick Start Ask the agent to check whether a skill for your intended task already exists and, if not, draft a new SKILL.md following this ecosystem's quality bar and package it into dist.
