← Back to archive

Repo of the Day

K-Dense-AI/scientific-agent-skills

Published: Oct 18, 2026

Open repository ↗

Turn any AI agent into an AI Scientist. The #1 Agent Skills library for science, used by 190,000+ scientists worldwide. 165 ready-to-use validated skills plus 100+ scientific databases covering biology, chemistry, medicine, and drug discovery. Compatible with Cursor, Claude Code, Codex, Pi, Antigravity, and the open Agent Skills standard

Summary

Scientific Agent Skills is a curated collection of 177 skill modules that extend compatible AI coding agents (Claude Code, Cursor, Codex, Google Antigravity) with documented procedures for working with scientific Python packages, public databases, and research platforms. Each skill ships as a SKILL.md plus optional code, references, and tests, so an agent can follow version-aware guidance for tasks like cheminformatics with RDKit, single-cell analysis with Scanpy, or querying sources like UniProt and ClinicalTrials.gov. The project is MIT licensed and accompanies the arXiv paper 2609.00065.

What it is useful for

  • Adding reliable, version-pinned procedures to an AI agent without writing each one yourself. Skills such as pathml, geopandas, and lab-hardware-cad ship specific documented baselines (e.g., PathML 3.0.8, GeoPandas 1.2.0, build123d 0.13.0).
  • Multi-step research workflows that span Python libraries and external data sources. Examples documented in the README include preparing bulk RNA-seq counts for handoff to PyDESeq2 and then pathway enrichment, validating cryo-EM reconstructions with RELION half-map checks, and running marine carbonate speciation with measurement uncertainty.
  • Lab platform integration through dedicated skills for Benchling, DNAnexus, LatchBio, Protocols.io, OMERO, Open Notebook, LabArchives, and Opentrons.
  • Scientific communication work, including evidence-traceable literature reviews, calibrated 1D NMR processing via nmrglue, and publication-quality figures.

How engineers can use it

The README documents several installation paths. For supported Agent Skills hosts:

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

With the GitHub CLI (v2.90.0+):

gh skill install K-Dense-AI/scientific-agent-skills
gh skill install K-Dense-AI/scientific-agent-skills scanpy
gh skill install K-Dense-AI/scientific-agent-skills --pin v2.71.0

For Cursor, symlink the checkout into the local plugins directory and reload the window:

mkdir -p ~/.cursor/plugins/local
ln -s "$(pwd)" ~/.cursor/plugins/local/scientific-agent-skills

On hosts that scan ~/.agents/skills/ or .agents/skills/, clone the repo into that path. Each SKILL.md lists its own compatibility and validation scope, so check it before depending on an older example.

A few limitations worth flagging. The healthcare, preclinical, and regulatory skills are described in the README as aids only; they are explicitly not certification, accreditation, method-release, or patient-specific diagnostic decisions. The collection no longer bundles docx/pdf/pptx/xlsx skills (removed in 2.72.0), so install those separately from Anthropics/skills if needed. Repository CI runs structural checks only; scientific dependencies are tested in separate environments described in the Testing section.