GitHub repo: 450+ AI medical skills
A new open GitHub project, AIPOCH, bundles more than 450 modular AI ‘skills’ for medical research workflows — from evidence synthesis to analysis and writing — aiming to bridge AI and research practice. The toolkit is presented as modular building blocks researchers can adapt to different study tasks. (x.com)
Most medical research does not happen in one giant “AI doctor” prompt. It happens as dozens of small jobs: search papers, extract methods, check guidelines, clean data, run an analysis, then draft a section of a manuscript. (github.com) AIPOCH is built around that reality. Its GitHub repository lists 450+ separate medical research “skills,” so a researcher can call a narrow tool for one step instead of asking one chatbot to do the whole project at once. (github.com) The basic idea is modular software. A skill is like a kitchen appliance with one job — blender, toaster, scale — and AIPOCH groups those one-job tools into medical research tasks. (github.com) The repository sorts those tasks into four buckets: Evidence Insights, Protocol Design, Data Analysis, and Academic Writing. Those map onto the usual path from reading the literature to submitting a paper. (github.com) Inside the Evidence Insights bucket, the repo includes skills with names like “PubMed search specialist,” “citation network mapping,” “retraction watcher,” and “open access scout.” That means the system is aimed at the messy front end of research, where people are still figuring out what is already known. (github.com) Inside Protocol Design, it includes tools such as “electronic case report form designer,” “institutional review board application assistant,” and “animal research reporting guideline architect.” Those are the forms and plans that usually slow a project down before any data gets analyzed. (github.com) Inside Data Analysis, the skills list reaches into actual study mechanics with items like survival analysis helpers, block randomization, pharmacokinetic and pharmacodynamic analysis, and quality-control checks. That pushes the project beyond paper summaries and into work researchers usually do in statistics software or scripts. (github.com) Inside Academic Writing, the repo includes manuscript-oriented tools such as method writing, journal matching, prior author letter drafting, and plagiarism prescreening. In other words, it is trying to cover the last mile too, not just the search box at the start. (github.com) The other notable piece is quality control. AIPOCH says it ships with a “Medical Skill Auditor,” which is a framework for checking a skill before it is deployed, instead of treating every prompt template as equally trustworthy. (github.com) This is also not locked to one interface. The GitHub page says the skills are built to work with OpenClaw and other agent platforms, including OpenCode and Claude, so the bet here is on reusable building blocks rather than one branded app. (github.com) The catch is that an open repository of 450+ skills is not the same thing as 450+ validated clinical tools. The public materials describe a research workflow toolkit, and any claim that a given skill is accurate enough for real medical decisions still has to be tested skill by skill. (github.com; dev.to) So the news here is less “someone built one super-medical model” and more “someone open-sourced a parts catalog for medical research agents.” If that approach works, the useful unit of medical artificial intelligence may end up being the small, auditable research step, not the giant all-purpose assistant. (github.com; news.ycombinator.com)
