177 validated skills for science and research: single-cell RNA-seq with Scanpy, molecular work with RDKit, differential expression, docking, microbiome pipelines, scientific writing with traceable evidence, plus lookups across 100+ databases (PubChem, ChEMBL, UniProt, ClinicalTrials.gov) and 70+ tuned Python package skills.
License: MIT · Created by K-Dense · The project lists support for Claude Code, Cursor, Codex, Gemini CLI, Antigravity, Pi
Free
What you get
177 validated skills for science and research: single-cell RNA-seq with Scanpy, molecular work with RDKit, differential expression, docking, microbiome pipelines, scientific writing with traceable evidence, plus lookups across 100+ databases (PubChem, ChEMBL, UniProt, ClinicalTrials.gov) and 70+ tuned Python package skills.
Formerly claude-scientific-skills.
Requirements
Python environment for the package skills
Install the Scientific Agent Skills skill
A community skill: it installs from its own repository and your agent runs the project's instructions locally with your own model. SkillGild lists it and credits the author; nothing runs on SkillGild.
With the SkillGild CLI (needs git). It clones K-Dense-AI/scientific-agent-skills and copies the skill folder into the right place for your agent; use --agent codex, cursor or gemini-cli for other clients.
gh skill install K-Dense-AI/scientific-agent-skills, or clone into ~/.claude/skills/
Or copy the folder yourself from K-Dense-AI/scientific-agent-skills into ~/.claude/skills/scientific-agent-skills/ for Claude Code, ~/.agents/skills/ for Codex, ~/.cursor/skills/ for Cursor or ~/.gemini/skills/ for Gemini CLI, then start a new session.
Try asking
Requests that trigger this skill once it is installed in your agent.
Run differential expression on these counts with PyDESeq2
Look up ChEMBL activity data for this compound
Write the methods section with traceable citations
Questions
How do I install the Scientific Agent Skills skill in Claude Code?
Run skillgild install scientific-agent-skills --agent claude-code: the SkillGild CLI clones K-Dense-AI/scientific-agent-skills and copies the skill folder into ~/.claude/skills/scientific-agent-skills. You can also use the project's own commands shown above, or copy the folder yourself. Use --agent codex, cursor or gemini-cli for other clients.
Is Scientific Agent Skills free and open source?
Yes. The upstream repository is published under the MIT license by K-Dense. SkillGild lists and credits it and installs it from source; it does not host or run it.
Which agents does Scientific Agent Skills work with?
The project lists Claude Code, Cursor, Codex, Gemini CLI, Antigravity, Pi. It is a plain Agent Skills folder (SKILL.md plus supporting files), so any client that reads that format can load it.
Does Scientific Agent Skills run on SkillGild?
No. Scientific Agent Skills is a community skill: your agent runs its instructions locally with your own model, and SkillGild lists, credits and installs it. Hosted SkillGild skills are marked differently and run through the SkillGild MCP server.
What does Scientific Agent Skills need to work?
Python environment for the package skills. Everything else is the usual: a coding agent that loads skills and the folder installed where that agent looks for it.
Related guides
Setup steps and worked examples for this kind of skill.
Publication-quality figures for ML papers: picks the chart type for your results, writes the matplotlib or seaborn code and checks its colors for contrast. Use it to compare methods, baselines or ablations in a chart, to check whether text and figure colors meet a contrast ratio, or to plan an architecture diagram from your method section.
An eight-phase research pipeline for Claude Code (scope, plan, retrieve with parallel search agents, triangulate, refine the outline, synthesise, critique with loop-back, package) that produces citation-backed reports with source-credibility scoring and automated validation.
Researches any topic across Reddit, X, YouTube, Hacker News, Polymarket, GitHub and the web for the last 30 days, ranks what people actually upvoted and shared, and synthesises a grounded summary or copy-paste-ready prompts.
Karpathy's observations on how LLMs fail at coding, codified into four rules the agent follows: think before coding (state assumptions, ask instead of guessing), simplicity first, surgical changes, and goal-driven execution.
Builds an animation from scratch the way a design engineer would: it chooses the curve, duration and properties, keeps motion off the main thread where it matters, and explains the choice.
Translates Apple's interface and motion principles into web code: spring curves, gesture physics, translucent materials and typography, as explained in Apple's WWDC design sessions.
Turn the project you just built into a short, shareable launch video with one command. Your agent reads the project code, so no live URL or screenshots are needed.
Anthropic's canvas-design skill guides Claude to produce finished visual artefacts (posters, covers, social graphics) by composing on an HTML canvas with deliberate typography, grid and colour decisions rather than defaulting to a template.
Anthropic's financial-services marketplace: a core financial-analysis plugin (comps, DCF, LBO, three-statement models in Excel, deck QC) plus vertical bundles for investment banking, equity research, private equity, fund administration, operations and financial advisors, 11 named agents and connectors to Daloopa, Morningstar, S&P Global, FactSet, PitchBook and more.
A universal SEO skill for Claude Code: 26 sub-skills and 19 specialist sub-agents covering technical audits, content quality, schema, E-E-A-T, local and e-commerce SEO, hreflang, backlinks, sitemaps, images, generative-engine optimisation and agent readiness (llms.txt, WebMCP).