Scientific Agent Skills
Research147 scientific research skills: genomics, drug discovery, proteomics, clinical ML, 100+ databases, for any Agent Skills agent.
Install
$ npx skills add K-Dense-AI/scientific-agent-skills
Update
$ gh skill update --all
Re-run `npx skills add` to refresh, or use `gh skill update --all` (GitHub CLI v2.90.0+). Install a topical subset, not all 147 skills — they add a lot of standing context.
When to use it
- You want your agent to run multi-step scientific pipelines — e.g. query ChEMBL for EGFR inhibitors, run RDKit SAR, dock with DiffDock, then write a report.
- You need reliable, provenance-rich access to scientific/clinical databases (PubChem, UniProt, COSMIC, ClinicalTrials.gov) instead of letting the agent improvise API calls.
- You work in bioinformatics, cheminformatics, proteomics, clinical ML, materials science, or geospatial analysis and want curated package skills (Scanpy, PyDESeq2, RDKit, PyTorch Lightning).
- You want literature review, scientific writing, lab automation (Benchling, Opentrons), or figure/slide generation driven by your agent.
Commands & flags
- npx skills add K-Dense-AI/scientific-agent-skills
- Install the full collection across Claude Code, Cursor, Codex, Google Antigravity, and other Agent Skills hosts.
- gh skill install K-Dense-AI/scientific-agent-skills scanpy
- Install a single skill (e.g. scanpy) via GitHub CLI instead of the whole bundle.
- gh skill update --all
- Update all installed skills to the latest published versions (GitHub CLI v2.90.0+).
- database-lookup
- Unified skill for deterministic, provenance-rich REST access to 78 public databases (PubChem, ChEMBL, UniProt, COSMIC, ClinicalTrials.gov).
- rdkit / scanpy / pydeseq2 / diffdock
- Per-package skills with curated docs and examples for cheminformatics, single-cell, differential expression, and docking workflows.
- gh skill install K-Dense-AI/scientific-agent-skills --pin v1.0.0
- Pin to a release tag for reproducible installs.
How to use it
Install via the Agent Skills standard (works in Claude Code, Cursor, Codex, Google Antigravity, and more):
npx skills add K-Dense-AI/scientific-agent-skills
Or pick individual skills with GitHub CLI (v2.90.0+) to avoid pulling all 147 into context:
gh skill install K-Dense-AI/scientific-agent-skills scanpy
gh skill update --all
Skills require uv to install their Python dependencies:
curl -LsSf https://astral.sh/uv/install.sh | sh
Once installed, the agent auto-discovers relevant skills; you can also name one directly in your prompt (e.g. "use database-lookup and rdkit"). Review each SKILL.md before installing — skills execute code and make network requests.
Why it earns its place
One of the largest curated open-source scientific skill collections for AI agents — 147 skills spanning genomics, drug discovery, proteomics, clinical ML, materials science, and scientific writing, plus a unified database-lookup skill covering 78 public databases with auditable provenance. Maintained by K-Dense Inc. (k-dense.ai), MIT-licensed, with all skills security-scanned approximately weekly via the Cisco AI Defense Skill Scanner, and built on the open Agent Skills standard so it works beyond Claude. It turns a coding agent into a research assistant that runs real end-to-end lab and analysis workflows from a single prompt.
Try it
> Use available skills: query ChEMBL for EGFR inhibitors with IC50 < 50nM, run SAR analysis with RDKit, dock candidates against the AlphaFold EGFR structure with DiffDock, then write a report.