What it does / problem it solves: Evidentia catches AI-fabricated medical citations before publication — it extracts DOIs/PMIDs/NCT/arXiv IDs from a document and resolves each against CrossRef, PubMed, and OpenAlex (no API key required), classifying every citation as Verified / Bibliographic mismatch / Hallucination / Content-review-needed, and distinguishes "real paper, wrong DOI" from "paper does not exist" via title fallback. It ships as a CLI (npx evidentia check file.md), an MCP server, and a Claude Code plugin/skill that adds a 15-criteria evidence appraisal with an A–F report (MIT, v1.0.0, deterministic engine with zero runtime deps, 38 unit tests + 17-case live benchmark).
Who uses this workflow / attribution: Built and used daily by a board-certified pediatrician (@kgraph57) for medical writing and fact-checking, based on a real use case — verifying citations in AI-assisted drafts before publishing.
Why this category: Added under Communication & Writing in alphabetical order (between Content Research Writer and family-history-research), since it is a writing/publishing workflow tool that complements the citation-adding skills already listed there. Only README.md was changed, matching the existing entry format.
🤖 Generated with Claude Code