What problem does it solve? Academic literature APIs fail silently with HTTP 200 responses that contain errors, empty bodies, or wrong records, producing confident but incorrect answers. This Skill turns a literature question into a bounded, rate-limited, reproducible retrieval across 11 scholarly databases with full provenance. ## Core Features & Use Cases - Multi-database search: Query PubMed, PMC, Europe PMC, bioRxiv, medRxiv, arXiv, OpenAlex, Crossref, Semantic Scholar, CORE, and Unpaywall with a selection guide that routes each intent to the right database. - Silent-failure detection: Bundled Python scripts catch HTTP-200 failure modes such as bodyless JATS XML, arXiv Error entries, and out-of-step pagination cursors, exiting non-zero instead of reporting wrong results. - Reproducible provenance: Every result includes endpoints, parameters, identifier conversions, count reconciliation, and access date so a human or agent can repeat the retrieval. - Use Case: Ask for all papers citing a specific DOI; the Skill queries Semantic Scholar and OpenAlex, paginates deterministically, reconciles retrieved counts against reported totals, and returns an auditable citation list. ## Quick Start Ask the assistant to find recent papers on CRISPR gene therapy and include open-access PDF links for each result.
