Which AI agents work with Scientific Agent Skills?

Any agent host supporting the open Agent Skills standard, including Claude Code, Cursor, Codex, Gemini CLI, Google Antigravity, and Claude Cowork. Community hosts like OpenClaw, NemoClaw, Pi, and Hermes are also supported via manual install.

Is this repository free and self-hostable?

Yes. It's MIT licensed and consists of static skill files (SKILL.md plus examples) that you install locally into your agent's skill directory. No server or hosted service is required.

Do I need to install extra software to use these skills?

You need an Agent Skills-compatible host and, for many skills, the underlying Python packages (RDKit, Scanpy, BioPython, etc.) that the skill documents. The skills themselves are documentation and code examples, not a runtime.

Can these skills make clinical or regulatory decisions on their own?

No. The README states that clinical, healthcare AI, imaging, and regulatory skills produce draft, evidence-preparation artifacts for qualified human review, not diagnosis, treatment, alarms, deployment, or certification decisions.

How do I keep installed skills up to date?

If you installed via GitHub CLI's gh skill, run `gh skill update` to check for updates or `gh skill update --all` to update everything. You can also pin to a specific release tag or commit SHA for reproducible installs.

How does this differ from just letting an agent call any Python package directly?

Agents can already call arbitrary Python packages or APIs on their own. These skills add curated, version-aware documentation, examples, and safety notes for specific packages and databases, making agent behavior more consistent and reliable for those workflows.

Scientific Agent Skills: 156 curated Agent

Scientific Agent Skills is a collection of 156 ready-to-use skill definitions that extend AI coding agents (Claude Code, Cursor, Codex, Google Antigravity, and other hosts supporting the open Agent Skills standard) into research assistants for science. Each skill packages documentation, code examples, and best practices for a specific scientific library, database, or workflow. It's built for computational biologists, cheminformaticians, data scientists, and research engineers who already use an AI coding agent and want it to handle domain-specific tasks reliably instead of guessing API syntax from scratch.

The skills span genomics, single-cell analysis, drug discovery, proteomics, clinical research, medical imaging, materials science, geospatial analysis, and lab automation. Rather than replacing what the agent can already do with any Python package or API, these skills give it curated, version-aware guidance for the tools and databases scientists use most.

Key features

156 packaged skills: covers genomics, cheminformatics, proteomics, clinical research, medical imaging, physics, materials science, geospatial science, and lab automation.
100+ database access: a unified database-lookup skill reaches 78 public databases (PubChem, ChEMBL, UniProt, COSMIC, ClinicalTrials.gov, and more), plus dedicated skills for DepMap, Imaging Data Commons, PrimeKG, and others.
70+ Python package skills: version-scoped workflows for RDKit, Scanpy, PyTorch Lightning, scikit-learn, PyTDC, pydicom, GeoPandas, pymatgen, Qiskit, and molecular dynamics tools like OpenMM and MDAnalysis.
Lab and platform integrations: skills for Benchling, DNAnexus, LatchBio, OMERO, Protocols.io, LabArchives, and Opentrons for connecting agents to lab systems.
Communication and methodology skills: literature review, evidence-traceable writing, grant writing, hypothesis generation, and regulatory documentation prep (ISO, ICH, USP, CLSI frameworks).
Consistent structure: every skill ships a SKILL.md with documentation, code examples, use cases, and reference material.
Multi-host compatibility: installable via npx, GitHub CLI's gh skill, or manual git clone into agent-specific skill directories.

Ideal use cases

Use this if you run an AI coding agent for research tasks and want it to reliably drive specific scientific packages or databases: pulling variant data from ClinicalTrials.gov, running a Scanpy single-cell pipeline, screening compounds with RDKit, or querying UniProt without hand-writing boilerplate each time. It also fits labs that want a shared, version-pinned set of workflows so multiple team members get the same agent behavior across projects.

It's a good fit for exploratory data analysis, prototyping bioinformatics pipelines, drafting evidence-based research documents, and integrating agents with lab platforms like Benchling or Opentrons.

It is not a substitute for domain expertise or regulatory sign-off. The README is explicit that clinical, diagnostic, and regulatory skills produce draft artifacts for qualified human review, not certification, treatment decisions, or deployment-ready outputs. If you need a production diagnostic system, an accredited method validation, or patient-specific clinical decisions, this repo gives you research-support scaffolding, not a finished regulated product.

Installation

The most common path is the npx installer, which works with supported hosts like Claude Code, Codex, Gemini CLI, Google Antigravity, and Cursor:

npx skills add K-Dense-AI/scientific-agent-skills

If you use the GitHub CLI (v2.90.0+), install with gh skill:

# Browse and install interactively
gh skill install K-Dense-AI/scientific-agent-skills

# Install a specific skill directly
gh skill install K-Dense-AI/scientific-agent-skills scanpy

# Target a specific agent host
gh skill install K-Dense-AI/scientific-agent-skills --agent cursor

You can pin to a release tag or commit SHA for reproducible installs:

gh skill install K-Dense-AI/scientific-agent-skills --pin v2.59.0

Keep installed skills current with:

gh skill update --all

For hosts without a built-in installer (OpenClaw, NemoClaw, Pi, Hermes, and others), clone the repo directly into the host's skill directory:

git clone https://github.com/K-Dense-AI/scientific-agent-skills.git ~/.agents/skills/scientific-agent-skills   # user-level
git clone https://github.com/K-Dense-AI/scientific-agent-skills.git .agents/skills/scientific-agent-skills      # project-level

Confirm the exact install path and any host-specific frontmatter requirements against your agent's current documentation, since conventions vary across hosts.