What problem does it solve? Finding usable medical and research datasets across public platforms is error-prone: OpenML paginated responses get truncated, quality metrics arrive as strings, and many UCI/GitHub dataset mirrors silently return 404 pages with HTTP 200 status codes. This Skill encodes verified platform behaviors and fallback strategies so dataset searches return accurate, deduplicated results. ## Core Features & Use Cases - OpenML API Querying: Searches datasets via /api/v1/json/data/list/ with safe limit=50 + offset pagination, parses the data.dataset array using did identifiers, and converts string quality metrics to floats. - Source Availability Verification: Validates GitHub and remote mirrors by inspecting the first line of file content rather than trusting HTTP status codes, correctly detecting disguised 404 pages. - Synthetic Data Fallback: When all public sources are dead (as confirmed for the UCI stroke dataset), generates a reproducible synthetic dataset with random.seed(42) matching the known schema, clearly labeled as synthetic. - Use Case: A researcher needs a breast cancer dataset for a study. The Skill paginates OpenML results, filters by medical keywords, cross-references existing papers in outputs/papers/ to avoid duplicates, and returns candidate datasets with metadata and download URLs. ## Quick Start Search OpenML for medical datasets matching the keyword "breast cancer" using paginated queries and return deduplicated candidates with metadata.
