Medical Research Skills
A 554-skill library built exclusively for medical and biomedical researchers, covering evidence discovery, protocol design, omics analysis, clinical statistics, and manuscript writing — every skill audited before release.
Last reviewed by the paperbanana team on Jul 13, 2026
Install
git clone https://github.com/aipoch/medical-research-skills ~/.claude/skills/medical-research-skillsWhat is Medical Research Skills?
Medical Research Skills is a curated library from AIPOCH containing 554 (and growing) agent skills built specifically for medical and biomedical research, distinct from general-purpose academic skill collections that only bolt on a few clinical extras. Every skill is organized into five categories — Evidence Insight, Protocol Design, Data Analysis, Academic Writing, and a smaller "Other" bucket for general non-research skills — and is reviewed through the project’s own MedSkillAudit, a domain-specific audit framework for medical research agent skills, before it goes live. The library works with any SKILL.md-compatible agent platform, explicitly listing Claude Code, Codex, Open Code, Hermes Agent, and OpenClaw as supported hosts.
Browsing by research stage rather than category shows the depth: 79 skills for literature and evidence discovery (PubMed/multi-database search strategy design, structured paper reading, evidence-gap and translational-opportunity finding, citation and retraction tracking), 68 for study design and protocol development (hypothesis generation, sample-size and power planning, 30+ disease- and method-specific protocol planners covering everything from Mendelian randomization to FAERS pharmacovigilance studies, and PROSPERO/IRB/IACUC registration assistance), 52 for omics and bioinformatics analysis (single-cell RNA-seq via Scanpy, differential expression via PyDESeq2, GO/KEGG/GSEA pathway analysis, immune-infiltration deconvolution, genomics and sequence tools), and 61 for clinical research and meta-analysis (PICOS generation, 15+ meta-analysis figure types, survival analysis, ROC/nomogram/decision-curve diagnostic modeling, and the full set of risk-of-bias tools — ROB2, NOS, QUADAS-2, PROBAST, QUAPAS).
Manuscript Writing & Publication is the largest single stage at 66 skills, spanning section-by-section drafting for CONSORT/STROBE/PRISMA/TRIPOD/STARD/CARE-compliant manuscripts, target-journal matching, reporting-guideline compliance checking, peer-review response drafting, NIH Specific Aims and Biosketch builders, and dissemination formats (conference abstracts, poster storylines, lay summaries, press releases). A further 77 skills round out research workflow and lab management — reagent expiry tracking, reproducibility and QC checks, scientific visualization (heatmaps, volcano plots, multi-panel figure assembly), computational pathology, and document production — plus a smaller set for research documentation, data-privacy (HIPAA/PHI de-identification, DICOM anonymization), and medical education (USMLE-style case generation, OSCE virtual patients).
Core capabilities
Domain-specific skill audit (MedSkillAudit)
Every skill in the library is reviewed and evaluated through MedSkillAudit, a domain-specific audit framework for medical research agent skills, before it is published, rather than being adapted from generic templates.
Evidence discovery and gap identification (79 skills)
PubMed Boolean query building, multi-database search strategy design, preprint surveillance, structured close/extensive paper reading, evidence-audited research gap finding, citation network and retraction tracking.
Study protocol design across 30+ specialized planners (68 skills)
Hypothesis generation, sample-size/power planning, and disease- or method-specific protocol planners covering clinical cohorts, Mendelian randomization, FAERS pharmacovigilance, network toxicology, and multi-omics integration, plus PROSPERO/IRB/IACUC registration helpers.
Omics and bioinformatics pipelines (52 skills)
Single-cell RNA-seq (Scanpy, scVI-tools, spatial transcriptomics), bulk differential expression (PyDESeq2), pathway/network analysis (GO/KEGG/GSEA/WGCNA), immune infiltration deconvolution, and genomics/sequence tools built on Biopython, BLAST, and SAM/BAM/VCF handling.
Clinical statistics and meta-analysis (61 skills)
Full meta-analysis pipeline with 15+ figure types (forest, funnel, Baujat plots), survival and time-to-event analysis, diagnostic/predictive modeling (ROC, nomograms, LASSO, XGBoost), and the complete risk-of-bias toolset (ROB2, NOS, QUADAS-2, PROBAST, QUAPAS).
Reporting-guideline-compliant manuscript writing (66 skills)
Section writers aligned to CONSORT/STROBE/PRISMA/TRIPOD/STARD/CARE, target-journal matching, reference integrity checking, claim-strength calibration, and peer-review response builders following ICMJE/CRediT authorship conventions.
What you can use it for
Designing a systematic review protocol
Use the PROSPERO systematic review protocol writer and eligibility criteria generator from the evidence-discovery category, then move into the meta-analysis pipeline for PICOS generation and screening.
Running a single-cell RNA-seq analysis for a manuscript figure
Use the Scanpy QC-to-clustering pipeline and automated cell-type annotation skills, then hand results to the volcano-plot or heatmap-beautifier visualization skills for publication-ready figures.
Drafting an SCI manuscript that meets reporting standards
Use the methods-section writing assistant (CONSORT/STROBE/PRISMA-aware) and the reporting-guideline compliance checker to catch missing items before submission, then the target-journal matcher to pick a venue.
Building a clinical prediction model with proper validation
Use LASSO logistic regression with cross-validation or the ML modeling skill (random forest/SVM/XGBoost/LightGBM), then validate with the calibration-curve and decision-curve-analysis skills and an external validation skill.
Responding to peer review on a resubmission
Use the reviewer-comment response drafter and author-response builder for point-by-point replies, then the resubmission deadline tracker to plan phase-appropriate revision tasks.
How to get started
- 1
Clone the repository
Run `git clone https://github.com/aipoch/medical-research-skills ~/.claude/skills/medical-research-skills` to vendor the full 554-skill library.
- 2
Pick your host platform
The library works with any SKILL.md-compatible agent — Claude Code, Codex, Open Code, Hermes Agent, or OpenClaw. For OpenClaw specifically, a one-command installer script (`scientific-skills/scripts/openclaw-install.sh`) copies all SKILL.md folders into `~/.openclaw/skills/`, skipping ones already installed.
- 3
Browse by category or research stage
Use the five top-level categories (Evidence Insight, Protocol Design, Data Analysis, Academic Writing, Other) or the eight research-stage groupings in the README to find the specific skill you need rather than loading the whole set at once.
- 4
Restart your agent to pick up new skills
After installing, restart your gateway or coding-agent session (e.g. `openclaw gateway restart` for OpenClaw) so the new skills are discovered.
How it compares to similar skills
Medical Research Skills goes deeper into clinical and biomedical workflows than any general academic suite in this directory. If your research spans other disciplines too, pair it with a broader catalog.
Qinyan Academic Skills
Pick qinyan-academic-skills if you need bioinformatics and clinical skills alongside grant writing, presentations, and general literature tools in one broader 18-category catalog.
Auto Empirical Research Skills
Pick auto-empirical-research-skills if your medical research is primarily observational/econometric (e.g. health-policy causal inference) rather than clinical-trial or omics-focused.
Academic Research Skills
Pick academic-research-skills if you want a single research-to-publication pipeline with citation-integrity gates rather than a large menu of discipline-specific medical skills.