MedSci Skills

MedSci Skills is a clinical research Skill set by Yoojin Nam, MD (Aperivue). The README says it covers literature and references, study design, statistics, figures, manuscript drafting, reporting-guideline checks and journal submission, plus a lane for building and validating medical-imaging AI models. The README lists 54 active Skills, listed here as one set.

It's built for researchers: it works with study data, tables and manuscripts, and searches databases such as PubMed, CrossRef and OpenAlex. The author says the model-evaluation module should not be run on real patients. A de-identification script creates anonymized files, a mapping table and an audit log on your machine; the mapping keeps the original values, and the author says not to give it to the AI.

Who it's for:Clinical and medical engineering researchers who want AI help with literature searches, statistics and manuscript revisions.

Before you start: Terminal install needs Node 18+ and Python 3.9+, using npx medsci-skills install; there are also classroom installers for Windows and macOS. It works with Claude Code, Codex, Cursor and GitHub Copilot. The literature connectors call PubMed, CrossRef, DOI, OpenAlex, Unpaywall, Europe PMC and PMC, and some need a contact email. Whatever you share is processed by your host AI, so don't upload anything that could identify a patient. An optional update-reminder hook writes to ~/.claude/settings.json and checks GitHub for a new version every 24 hours; the classroom installer turns it on by default, while terminal installs make it opt-in. If you use these features, weigh the safety risks yourself. AI Inspiration World is not responsible for any consequences. The root license is MIT, but some reference attachments use non-open licenses such as CC BY-NC, and there are optional AGPL dependencies, so check each before commercial use. Mostly English, with Simplified Chinese and Korean READMEs linked.

Based on the project's public documentation. Not tested by AI Inspiration World.