About this project

Scientific Agent Skills is an open-source collection of 166 ready-to-use skills for AI agents that support the open Agent Skills standard. It is designed to give a general-purpose coding agent curated, version-aware guidance for scientific libraries, databases and workflows, instead of relying on the agent to rediscover APIs from scratch. What the repository contains: - 100+ scientific and financial database integrations. A unified database-lookup skill covers 78 public databases such as PubChem, ChEMBL, UniProt, COSMIC, ClinicalTrials.gov, FRED and USPTO, with dedicated skills for DepMap, Imaging Data Commons, PrimeKG, NCATS ARAX, U.S. Treasury Fiscal Data, Hugging Science, OneKGPd, Genomic Intelligence and AlphaGenome. Multi-database packages such as BioServices, BioPython and gget extend coverage further. - 70+ optimized Python package skills with explicit, version-aware workflows for libraries including RDKit, Scanpy, PyTorch Lightning, scikit-learn, PyTDC, PathML, pydicom, NeuroKit2, PufferLib, QuTiP, GeoPandas, pymatgen, BioPython, Qiskit, OpenMM and MDAnalysis. - 9 scientific integration skills for platforms such as Benchling, DNAnexus, LatchBio, OMERO, Protocols.io, Open Notebook, Ginkgo Cloud Lab, LabArchives and Opentrons. - 30+ analysis and communication tools covering literature review, evidence-traceable writing, peer review, document processing, full-text paper and regulatory filing retrieval, search integrations, PPTX posters and slides, schematics, infographics and Mermaid diagrams. - 10+ research and clinical tools including hypothesis generation, grant writing, aggregate clinical decision-support research, PK/PD modelling and simulation, BIDS, ISO standards-readiness evidence preparation, analytical method validation and transfer, scenario analysis and workflow-derived skill drafting. Each skill ships with a SKILL.md document, practical code examples, use cases, integration guidance and reference material. Skills that bundle scripts also have a test suite, and CI enforces structural contracts such as frontmatter, link resolution and script parsing. Installation options described in the README include the npx skills installer, the GitHub CLI gh skill command with version pinning and update support, and Agent Plugins packaging for clients such as Cursor and Codex. Manual installation into user-level or project-level skill directories is also documented, along with notes for hosts such as NemoClaw that restrict outbound networking. The README includes a prominent security disclaimer: skills can execute code, install packages, make network requests and modify files, so users are advised to install only what they need, read each SKILL.md before installing, check contribution history, run the Cisco AI Defense skill scanner locally, and report suspicious behavior. Security scans are run on a recurring schedule and results are published in the repository. The project is MIT licensed, cites an arXiv paper, and is maintained by K-Dense with community contributions. It is aimed at researchers, labs and engineers who want an AI agent to handle multi-step scientific workflows across biology, chemistry, medicine, materials science, physics, geospatial analysis and related fields.