Production-grade agent skills for biomedical NLP — clinical named entity recognition, de-identification, relation extraction, entity linking, and more. Built for reproducible biomedical NLP workflows.
Research use only: these skills support biomedical NLP workflows and are not medical advice or a substitute for clinical validation. Check source licenses, access terms, and de-identification requirements before using protected health information.
Many biomedical and medical agent-skill repositories cover a broad spectrum — genomics, drug discovery, imaging, and EHR integration. This repo focuses specifically on the NLP layer that connects language models to biomedical text.
git clone https://github.com/infonality/biomedical-nlp-skills.git ~/.claude/skills/biomedical-nlp-skillsgit clone https://github.com/infonality/biomedical-nlp-skills.git ~/.codex/skills/biomedical-nlp-skillsgemini skills install https://github.com/infonality/biomedical-nlp-skills.git --path skillsCopy skill folders to .cursor/skills/ in your project.
npx skills add --repo infonality/biomedical-nlp-skills| Skill | Description |
|---|---|
clinical-ner |
Extract diagnoses, medications, procedures, lab values, and anatomical sites from unstructured clinical text. scispaCy, Med7, ClinicalBERT, and rule-based pipelines with negation detection. |
de-identification |
De-identify clinical text per HIPAA Safe Harbor (18 categories). Regex, NER-based, hybrid, and Presidio pipelines with date shifting and i2b2/n2c2 evaluation. |
relation-extraction |
Drug-disease, drug-drug, gene-protein, and adverse event relations. Dependency-rule, transformer-based (BioBERT/ClinicalBERT), and LLM-prompted. ChemProt, BC5CDR, n2c2 ADE coverage. |
pico-extraction |
Population/Intervention/Comparison/Outcome frames from RCT abstracts and ClinicalTrials.gov. Rule-based, PubMedBERT, PIBOSO classification, and LLM zero-shot. EBM-NLP evaluation. |
entity-linking |
Normalize entities to UMLS CUIs, MeSH, SNOMED CT, RxNorm, ICD-10, HGNC, UniProt. scispaCy linker, string matching, semantic matching, cross-vocabulary mapping via UMLS. |
clinical-section-segmentation |
Segment clinical notes into standard sections (CC, HPI, PMH, Meds, PE, Assessment, Discharge). Regex, fuzzy header matching, and line-classification. MIMIC-III coverage. |
biomedical-qa |
Retrieval-augmented QA over PubMed and clinical notes. BM25, MedCPT dense, hybrid retrieval, cross-encoder reranking. BioASQ and PubMedQA evaluation. |
literature-triage |
Systematic review screening automation. Deduplication, rule-based pre-filtering, PubMedBERT classification, and active learning. Cochrane/RobotReviewer workflows. |
shared-task-replication |
Reproduce i2b2, n2c2, BioCreative, BioASQ, BioRED benchmarks. Data access, format conversion (stand-off → BIO, BioCreative → spaCy), evaluation scripts, and published baselines. |
Claude Code · Codex CLI · Gemini CLI · Cursor · OpenCode · GitHub Copilot · any agent supporting the Agent Skills spec
MIT