RepoPicks Radar

Updated every day around 23:30 UTC

Guides

Scientific Agent Skills

K-Dense-AI/scientific-agent-skillsReviewed by AI

Checked by RepoPicks Radar on

What is Scientific Agent Skills?

This is a library of 177 written skills you install into an AI assistant such as Claude Code, Cursor, Codex or Gemini CLI. Each one teaches the assistant how to use a specific scientific tool, Python package or public database correctly — gene data, molecules, mass spectrometry, clinical trial records, satellite images, statistics and figures — with the conventions and checks that field expects. The skills are only text and helper scripts: you still need Python 3.13, the uv package manager, a paid AI assistant and the scientific packages each skill uses.

How easy is it to install Scientific Agent Skills?

C · install 1 or 2 tools · macOS

  • Windows D
  • macOS C
  • Linux C

On macOS and Linux it is C: install Python 3.13 and uv, then one command such as `npx skills add K-Dense-AI/scientific-agent-skills` or `gh skill install`, and after that each skill still needs its own packages installed. Windows is D because the Prerequisites list it only "with WSL2", so you first set up a whole Linux environment inside Windows.

Scientific Agent Skills on GitHub

Stars
56,728on Nov 10, 2026
Downloads
—all releases
Growth
+755stars in 7 days

Stars over time

56,00056,20056,40056,60056,800 Nov 4Nov 6Nov 8Nov 10 56,00056,20056,40056,60056,800 Nov 4Nov 7Nov 10

Facts

Runs on
Windows, macOS, Linux
You need to install
Python
To build from source
Python
Disk space
100 MB–300 MB · estimate
Accounts or keys
The model provider behind your AI assistant (Anthropic, OpenAI, Google and others), Scientific and search services used by individual skills (Exa, Parallel, Benchling, NCBI, Materials Project, OpenRouter and more)
Cost
Free, API costs extra · The skills are free (MIT). The AI assistant that runs them is not: you need a paid plan or your own API keys, and some skills call services that charge or require registration.
License
MIT (Permissive)
Can replace
NotebookLM
Last change on GitHub
Nov 9, 2026

Scientific Agent Skills risk signals (9)

  • AttentionInstall command runs a script from the internet

    Install command pipes a remote script into a shell: curl -LsSf https://astral.sh/uv/install.sh | sh

    automatic check

  • Attentionskills can run code on your machine

    README.md Security Disclaimer: "Agent Skills are powerful — they can instruct your AI agent to run arbitrary code, install packages, make network requests, and modify files on your system. A malicious or poorly written skill has the potential to steer your coding agent into harmful behavior." The authors add that they "cannot guarantee that every skill has been exhaustively reviewed" and advise "Do not install everything at once."

    AI review

  • Attentioncommunity contributions less reviewed

    README.md: "Skills authored by K-Dense (K-Dense-AI) have been through our internal review process. Community-contributed skills have been reviewed to the best of our ability, but with limited resources", and "the repository now includes many community contributions that we may not have reviewed as thoroughly". Weekly automated scans are published to docs/security-report.md.

    AI review

  • Attentionmedical and clinical limits

    Several skill groups touch patient-related work and the README bounds them explicitly: "Healthcare AI & Biosignal Research - EHR and model research, physiological signal analysis, and retrospective validation—not patient-specific diagnosis, treatment, alarms, or deployment decisions", and clinical reports stay "visibly marked drafts" for qualified review. Nothing here is a medical device or a diagnosis.

    AI review

  • Infodependencies are not installed for you

    README.md Prerequisites: "Installing the skill files does not install their dependencies", and "Use a separate environment for each scientific workflow: some skills require different Python versions or incompatible package pins." Getting one workflow to run can mean a long package install.

    AI review

  • Infounverified popularity claim

    The repository description says "The #1 Agent Skills library for science, used by 250,000+ scientists worldwide" and "177 ready-to-use validated skills". No source is given for either figure, and the 2.71.0 note in the same README speaks of "all 181 skills".

    AI review

  • Infosome skills removed over licensing

    README.md "What's new in 2.72.0": the docx, pdf, pptx and xlsx skills "which were vendored from anthropics/skills under Anthropic's own license, are no longer bundled". The repository is MIT, but individual skills have carried other licences.

    AI review

  • Infopromotes the company products

    README.md opens with several promotions for the same company's other product (K-Dense BYOK), a webinar, social accounts and "Enterprise Ready - Commercial support available for advanced needs", and asks readers to star the repository.

    AI review

  • Infodeclared alternative false match

    The automatic check read "invideo-ai" out of the line "Open Notebook (self-hosted NotebookLM alternative — PDFs, videos, audio, web pages...)". That line is about one integration skill and about NotebookLM, not about a paid video product, so invideo-ai is not listed here.

    AI review

Evidence is shown in its original language.

What Radar could not check about Scientific Agent Skills

  • Whether a given skill still matches the current version of the package or database it documents
  • Whether the results of a workflow are scientifically correct — the skills guide the assistant, they do not check the science
  • How much disk space and how many packages one scientific workflow really needs
  • What each skill sends to outside services and which ones charge for it
  • How thoroughly any individual community-contributed skill was reviewed

Similar projects

The RepoPicks app

Coming soon: install in 1 click with the RepoPicks app.

Before you download

See the release page on GitHub