What problem does it solve? Scientific, regulatory, and financial facts often need to be retrieved reproducibly from authoritative databases rather than inferred from general knowledge, but each database has different endpoints, identifier formats, pagination rules, and rate limits. This Skill turns a user's intent into a bounded, rate-limited, fully documented API retrieval that another agent or human can repeat. ## Core Features & Use Cases - 78 documented databases: Covers physics, chemistry, genomics, clinical, patents, economics, and demographics, each with a reference file detailing endpoints, parameters, and worked queries. - Identifier resolution workflows: Converts between UniProt, Ensembl, NCBI Gene, PubChem CID, ChEMBL, rsID, MONDO, and other identifier systems when a query fails. - Reproducible retrieval contract: Counts first, paginates exhaustively, reconciles retrieved vs expected totals, and returns provenance including endpoints, parameters, access date, and warnings. - Use Case: Ask for all clinical trials for a given drug target; the Skill selects ClinicalTrials.gov, resolves the target identifier, paginates through results, and returns a structured table with count reconciliation and query provenance. ## Quick Start Use the database-lookup skill to retrieve all known binding affinities for UniProt target P35355 from BindingDB with full provenance.
