What problem does it solve? Finding usable public medical datasets is slow and error-prone: popular sources like UCI have removed classic datasets, GitHub mirrors are dead, and APIs like OpenML, Crossref, and PubMed have non-obvious parameter quirks that silently return empty results. This Skill encodes verified knowledge of which sources are accessible, which are dead, and how to call each API correctly. ## Core Features & Use Cases - Verified Source Registry: Distinguishes accessible datasets (e.g., OpenML Cardiovascular-Disease-dataset, DID=45547, 70,000 records) from unavailable ones (UCI stroke data, dead GitHub mirrors) with concrete 404/authentication evidence. - API Quirk Handling: Documents correct endpoint patterns, such as using OpenML /api/v1/json/data/list without limit/500, distinguishing list vs. detail response structures, and using query= with quote_plus() for Crossref and + for PubMed spaces. - Golden Test Set: Ships three weighted test cases covering the normal discovery path, the dead-link failure path, and API parameter traps, requiring evidence-backed conclusions. - Use Case: A medical AI researcher needs a cardiovascular dataset; the Skill queries OpenML, filters locally, and returns a verified dataset entry with name, source, URL, size, and access type instead of a dead UCI link. ## Quick Start Ask the agent to find a public dataset for a medical domain such as cardiovascular disease and return verified sources with access details.
