Scientific Agent Skills
A 148-skill library (as of July 2026) that turns any AI agent into a research assistant for biology, chemistry, medicine, and drug discovery — with unified access to 100+ scientific databases.
Last reviewed by the paperbanana team on Jul 13, 2026
Install
npx skills add K-Dense-AI/scientific-agent-skillsWhat is Scientific Agent Skills?
Scientific Agent Skills (formerly Claude Scientific Skills) is a collection of 148 ready-to-use skills (as of July 2026), maintained by K-Dense, that equips any AI coding agent with curated documentation, examples, and best practices for scientific computing. Instead of relying on an agent to guess how to call RDKit, Scanpy, or PyMC correctly, each skill packages a `SKILL.md` with practical code examples, use cases, and reference material for a specific package, database, or scientific workflow — spanning cancer genomics, drug-target binding, molecular dynamics, RNA velocity, geospatial science, time-series forecasting, and more.
The collection is explicitly cross-disciplinary: bioinformatics and genomics, cheminformatics and drug discovery, proteomics and mass spectrometry, clinical research and precision medicine, healthcare AI and clinical ML, medical imaging and digital pathology, machine learning, materials science and physics, engineering and simulation, data analysis and visualization, geospatial science, laboratory automation, scientific communication, multi-omics and systems biology, and protein engineering are all covered under one repository. A single unified database-lookup skill provides deterministic, provenance-rich access to 78 public databases — PubChem, ChEMBL, UniProt, COSMIC, ClinicalTrials.gov, FRED, USPTO, and more — while multi-database packages like BioServices, BioPython, and gget extend coverage past 100 databases in total.
Originally built for Claude Code, Scientific Agent Skills now follows the open Agent Skills standard, so the same skills work unmodified in Cursor, Codex, Gemini CLI, Google Antigravity, OpenClaw, Hermes, and Pi. The README frames the goal plainly: transform your AI coding agent into an "AI Scientist" on your own desktop, capable of executing multi-step scientific workflows — such as querying ChEMBL for inhibitors, running structure-activity analysis with RDKit, docking candidates with DiffDock, and cross-checking resistance mechanisms in PubMed — from a single prompt.
Core capabilities
148 curated scientific skills
Covers 70+ optimized Python package skills (RDKit, Scanpy, PyTorch Lightning, scikit-learn, BioPython, pyzotero, BioServices, PennyLane, Qiskit, OpenMM/MDAnalysis, scVelo, TimesFM, and more) plus 9 lab-platform integrations (Benchling, DNAnexus, LatchBio, OMERO, Protocols.io, Opentrons, and others).
Unified access to 100+ scientific databases
A single database-lookup skill covers 78 public databases with auditable filters and provenance (PubChem, ChEMBL, UniProt, PDB, AlphaFold, KEGG, Reactome, STRING, ClinVar, COSMIC, ClinicalTrials.gov, FDA, FRED, USPTO, SEC EDGAR, and more), plus dedicated skills for DepMap, Imaging Data Commons, PrimeKG, and Hugging Science.
Full drug-discovery and cheminformatics stack
RDKit and Datamol for molecular manipulation, DeepChem and TorchDrug for deep learning, DiffDock for docking, OpenMM + MDAnalysis for molecular dynamics, MedChem for drug-likeness, and PyTDC for benchmarks — chainable into an end-to-end virtual screening or lead-optimization pipeline.
Clinical and precision-medicine tooling
Skills for ClinicalTrials.gov, ClinVar, COSMIC, and ClinPGx variant interpretation, DepMap cancer dependency data, pydicom/histolab/PathML medical imaging, and clinical-report and treatment-plan generation for precision-therapeutics workflows.
Multi-platform, portable install
Installs via `npx skills add`, `gh skill install` (with version pinning to a release tag or commit SHA), or a direct git clone into the shared `~/.agents/skills/` convention — the same skills work across Cursor, Claude Code, Codex, Gemini CLI, Antigravity, OpenClaw, Hermes, and Pi.
Security-scanned contributions
Every skill runs through the Cisco AI Defense Skill Scanner (LLM-based behavioral scanning for prompt injection, data exfiltration, and malicious code) on an approximately weekly cadence, with results tracked in SECURITY.md.
What you can use it for
End-to-end drug discovery pipeline
Query ChEMBL for EGFR inhibitors under an IC50 threshold, analyze structure-activity relationships with RDKit, generate analogs with datamol, virtually screen against an AlphaFold structure with DiffDock, and cross-check resistance mechanisms in PubMed and COSMIC — using database-lookup, rdkit, datamol, diffdock, and paper-lookup skills together.
Single-cell RNA-seq analysis
Load 10X Genomics data with Scanpy, run QC and doublet removal, integrate with Cellxgene Census, identify cell types via NCBI Gene markers, run differential expression with PyDESeq2, and infer gene regulatory networks with Arboreto.
Multi-omics biomarker discovery
Analyze RNA-seq with PyDESeq2, process mass spec with pyOpenMS, integrate metabolites from HMDB/Metabolomics Workbench, map proteins to pathways via UniProt/KEGG, correlate omics layers with statsmodels, and search ClinicalTrials.gov for relevant studies.
Clinical variant interpretation from a VCF file
Parse a VCF with pysam, annotate variants with Ensembl VEP, query ClinVar for pathogenicity and COSMIC for cancer mutations, check ClinPGx for pharmacogenomics, and generate a clinical report with the docx/pdf document-processing skills.
Systems biology network analysis
Query NCBI Gene for annotations, retrieve sequences from UniProt, identify interactions via STRING, map to Reactome/KEGG pathways, reconstruct gene regulatory networks with Arboreto, and assess druggability with Open Targets.
How to get started
- 1
Install uv and pick an install method
Install the uv Python package manager (required for skill dependencies), then install the skills with `npx skills add K-Dense-AI/scientific-agent-skills` — the official cross-platform method that works for Claude Code, Cursor, Codex, Gemini CLI, and Google Antigravity alike.
- 2
Or install selectively with gh skill
If you use GitHub CLI v2.90.0+, run `gh skill install K-Dense-AI/scientific-agent-skills` to browse interactively, or target one skill directly, e.g. `gh skill install K-Dense-AI/scientific-agent-skills scanpy`. You can pin to a release tag or commit SHA for reproducible installs.
- 3
Install only what you need
The README explicitly recommends not installing the full 148-skill collection at once — read each `SKILL.md` first, and prefer a topical subset relevant to your domain (e.g. cheminformatics or clinical research) to keep standing context and review surface small.
- 4
Prompt your agent to use available skills
Once installed, the agent auto-discovers relevant skills; you can also invoke one explicitly by naming it in your prompt (e.g. "use the database-lookup and rdkit skills to..."). Run `skill-scanner scan` yourself first if you are installing third-party or community-contributed skills you have not reviewed.
How it compares to similar skills
Scientific Agent Skills is a domain-tooling library for wet-lab and computational science rather than a manuscript-writing pipeline. If your task is closer to drafting or reviewing a paper than running scientific computation, a writing-focused suite in this directory may be a better fit.
Academic Research Skills
Pick Academic Research Skills if your priority is a citation-integrity research-to-write-to-review pipeline for a manuscript, rather than domain-specific scientific computing and database tooling.
Medical Research Skills
Pick medical-research-skills if you want a suite scoped tightly to clinical and medical research workflows rather than the full cross-discipline science collection covering chemistry, physics, and engineering too.
AI Research Skills
Pick ai-research-skills if your work centers on ML-research-paper output rather than wet-lab or computational-science tooling across many scientific disciplines.