What problem does it solve? Retrieving facts from scientific, regulatory, financial, and demographic databases often fails due to wrong identifier formats, incomplete pagination, undocumented rate limits, and missing provenance. This Skill turns a user's intent into a reproducible, auditable API retrieval across 78 documented public databases instead of relying on inferred general knowledge. ## Core Features & Use Cases - Guided database selection: A selection guide maps user intent to the authoritative database across domains like genomics, chemistry, clinical trials, patents, economics, and demographics, with per-database reference files covering endpoints, parameters, and worked examples. - Reproducible retrieval workflow: Enforces a retrieval contract with explicit filters, identifier conversion (e.g., gene symbol to NCBI Gene ID to Ensembl ID), count-first pagination, and count reconciliation so results can be repeated by another agent or human. - Safe API execution: Handles POST-only GraphQL APIs, API key discovery without exposing secrets, rate-limit serialization, URL encoding of special characters, and treats all API responses as untrusted third-party data. - Use Case: A researcher asks for all clinical trials and known binding affinities for a compound. The Skill resolves the compound name to a PubChem CID, queries ClinicalTrials.gov and BindingDB with bounded paginated calls, and returns a structured result table with endpoints, access dates, and count reconciliation. ## Quick Start Ask the assistant to look up a specific fact from a named public database, such as retrieving all UniProt entries for a given human gene with full provenance.
