What Are K-Dense Scientific Skills?
K-Dense Scientific Skills is a collection of 170+ ready-to-use agent skills spanning 17 scientific domains. Each skill is a self-contained SKILL.md file that transforms Claude Code, Cursor, or OpenClaw into a domain-specific research assistant — connected to 250+ scientific databases including PubMed, UniProt, ChEMBL, RCSB PDB, and many more.
With 15K+ GitHub stars, it's one of the most popular scientific skill libraries in the AI agent ecosystem.
The 17 Scientific Domains
| Domain | Example Skills |
|---|---|
| Bioinformatics & Genomics | BLAST search, variant annotation, pathway analysis |
| Cheminformatics & Drug Discovery | Molecular docking, ADMET prediction, compound similarity |
| Proteomics | Protein structure prediction, mass spec analysis |
| Clinical Research | Trial design, patient cohort analysis, outcome prediction |
| Healthcare AI | Medical image analysis, clinical NLP |
| Medical Imaging | DICOM processing, segmentation, radiomics |
| ML / AI | Model training, hyperparameter tuning, evaluation |
| Materials Science | Crystal structure prediction, property calculation |
| Physics & Astronomy | Spectral analysis, cosmological simulation |
| Engineering & Simulation | FEA, CFD, circuit analysis |
| Data Analysis | Statistical testing, visualization, time series |
| Geospatial Science | GIS analysis, remote sensing |
| Lab Automation | Protocol generation, instrument control |
| Scientific Communication | Paper writing, figure generation |
| Multi-omics | Cross-omics integration, pathway enrichment |
| Protein Engineering | Directed evolution, stability prediction |
| Research Methodology | Experimental design, power analysis |
Installation
For Claude Code
Clone the repository and point Claude Code to it:
git clone https://github.com/K-Dense-AI/claude-scientific-skills.git
cd claude-scientific-skillsSkills are in the skills/ directory. Claude Code automatically discovers SKILL.md files when you work within the project.
For OpenClaw / NanoClaw
Skills follow the standard SKILL.md format and are compatible with any OpenClaw-based agent. Copy individual skill directories to your agent's skills folder:
cp -r skills/bioinformatics-genomics ~/.openclaw/skills/Example: Running a Drug Discovery Workflow
Here's what a typical interaction looks like after installing the cheminformatics skills:
You: Search ChEMBL for compounds active against EGFR with IC50 < 100nM,
then predict ADMET properties for the top 5 hits.
Agent: I'll use the ChEMBL search skill to find active compounds...
[Searches ChEMBL API for EGFR inhibitors]
Found 847 compounds. Filtering by IC50 < 100nM: 142 hits.
Top 5 by potency:
1. Osimertinib (IC50: 3.2nM) ...
Running ADMET prediction for these 5 compounds...
[Uses RDKit + ML models for property prediction]Connected Databases
The skills connect to 250+ scientific databases and APIs. Here are some highlights:
- Life Sciences: PubMed, UniProt, RCSB PDB, Ensembl, KEGG, Reactome
- Chemistry: ChEMBL, PubChem, DrugBank, ZINC
- Clinical: ClinicalTrials.gov, FDA, WHO
- Genomics: NCBI, dbSNP, ClinVar, gnomAD
- General: arXiv, Semantic Scholar, CrossRef