Scientific Agent Skills
Research146 skills giving agents procedural access to bioinformatics, ML, and 78+ research databases.
Install
$ npx skills add K-Dense-AI/scientific-agent-skills
Update
$ npx skills add K-Dense-AI/scientific-agent-skills
Re-running fetches the latest version.
When to use it
- Running computational science pipelines (pathology WSI, single-cell, RL training) without hand-writing boilerplate
- Querying public scientific databases (PubChem, ChEMBL, UniProt, PrimeKG, ClinicalTrials.gov, and 70+ more)
- Autonomous hypothesis generation and drug-repurposing over knowledge graphs
- Deep literature research that prioritizes preprints (arXiv, bioRxiv) over commercial sources
Commands & flags
- npx skills add K-Dense-AI/scientific-agent-skills
- Install the full skill bundle via the open skills CLI (Claude Code, Codex, Gemini CLI, Cursor, and more).
- npx skills add <skill-name>
- Install a single skill by name instead of the whole bundle.
- pathml
- WSI pathology pipeline: load -> Macenko/Vahadane normalization -> tissue detection -> feature extraction.
- stable-baselines3
- RL training skill that forces observation spaces to np.uint8 in [0, 255].
- primekg
- Query the PrimeKG knowledge graph for drug-target interactions and repurposing candidates.
How to use it
Install with the open skills CLI (works across Claude Code, Codex, Gemini CLI, Cursor, and more), or install a single skill by name.
npx skills add K-Dense-AI/scientific-agent-skills
Each skill's SKILL.md encodes a strict order of operations to prevent hallucinated parameters — e.g. pathml dictates load → Macenko/Vahadane normalization → tissue detection → feature extraction; stable-baselines3 forces observation spaces to np.uint8 in [0, 255].
Why it earns its place
It dissolves the integration friction that — more than raw model intelligence — bottlenecks computational science. By shipping pre-configured schemas, database access, and deterministic procedures for highly specialized domains, a generalist agent can formulate and test scientific hypotheses against real data with far less setup and far fewer syntactic hallucinations.
Try it
> Use the primekg skill to find drug-target interactions for repurposing against a disease gene