What problem does it solve? AI agents can build almost anything on request, so teams skip the question "has someone already solved this?" and pay for it in tokens, iteration rounds, and unmaintained hand-rolled code. This Skill reinstates that check and answers it honestly, verifying candidates against source rather than recommending from memory. ## Core Features & Use Cases - Prior-art ladder: Walks rungs from stdlib and already-installed dependencies up through mature libraries, frameworks, mined repo patterns, and building from scratch, ordered by integration cost. - Search and vet tooling: Uses search.py (GitHub repo/code search via gh), fetch_repo.py (safe shallow clones into a marked cache), and vet.py (license detection plus OSV vulnerability checks with honest CONFIRMED/INDICATIVE labeling). - Divert test and red team: Challenges every candidate on coverage, maintenance, and reversibility, then dispatches an independent adversarial review before recommending adoption. - Use Case: Before adding rate limiting to a FastAPI service, the Skill checks installed dependencies and deployed infrastructure first, searches GitHub with multiple query phrasings, vets survivors for license and CVEs, and presents a gated recommendation for approval. ## Quick Start Before we build this feature, check whether a maintained library, an existing dependency, or a well-tested open-source pattern already solves it.