What problem does it solve? Choosing which agent skills to add to a dotfiles environment is ad-hoc and risky: skills consume context, may overlap with existing ones, and come from sources of varying quality. This meta-skill provides a disciplined find-evaluate-adopt pipeline so skills are added deliberately rather than impulsively. ## Core Features & Use Cases - Catalog-first lookup: Checks a curated, pre-vetted catalog (references/catalog.md) matched to the environment's stack (Astro/Svelte/TS, C/C++, Python, CTF/security) before searching external sources. - Tiered cross-source search: Searches prioritized GitHub sources (anthropics/skills, skill registries, awesome-lists) only when the catalog misses, with a 7-axis rubric covering fit, non-redundancy, maintenance, license, frontmatter, body quality, and footprint. - Mandatory eval gate and vendored adoption: Requires waxa eval convergence before adoption, then pins accepted skills by commit SHA in coding-agents/vendor/manifest.tsv instead of using APM. - Use Case: When you ask "is there a skill for Playwright testing?", it checks the catalog, proposes the vetted candidate with its install line, and stops—avoiding unvetted wild searches. ## Quick Start Ask the agent to find and evaluate a skill for a specific task, for example whether a Playwright testing skill should be adopted into this environment.
