jimliu/science-skills — explained in plain English
Analysis updated 2026-05-18
Configure an AI assistant to run protein structure prediction workflows like AlphaFold2.
Automate literature review or figure composition for a scientific paper.
Add an MCP server that gives an assistant access to biomedical data or chemistry tools.
Build a specialized onboarding or transcript review agent for a science assistant.
| jimliu/science-skills | keyai/calculusmadeeasy-zh | trading-2028/polymarket-ai-trading | |
|---|---|---|---|
| Stars | 216 | 215 | 219 |
| Language | HTML | HTML | HTML |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 3/5 | 1/5 | 3/5 |
| Audience | researcher | writer | developer |
Figures from each repo's GitHub metadata at analysis time.
These are configuration and prompt assets meant to run inside the Claude Science assistant, not a standalone application.
Science Skills is a collection of configuration files that power Claude Science, an AI assistant made for scientific computing tasks. Rather than being a standalone app you run, this repository holds the building blocks the assistant uses: a base system prompt describing how it should behave, individual skills that teach it specific scientific workflows, agent profiles for specialized roles, and MCP servers, which are small connector programs that give the assistant access to external scientific tools and data. The skills cover a wide range of scientific work. Several focus on biomolecular modeling, including tools like AlphaFold2, Boltz, ESMFold2, and ProteinMPNN, which are used to predict or design protein structures. Others handle scientific analysis and reporting tasks such as reviewing literature, exploring PDF documents, and composing figures for papers. There are also skills for working with single cell biology and genomics data, skills for setting up remote computing environments, and skills related to customizing the assistant itself. Each skill is organized around a file called SKILL.md, which starts with a short block describing the skill's name, description, and license, followed by detailed instructions covering how to carry out the task, what inputs are expected, what can go wrong, and how to check the result. Agent profiles work similarly but describe specialized roles, such as a general purpose science agent, an onboarding helper for new users, a reviewer that checks another agent's work for unsupported claims, and a bookmarking agent that flags useful parts of a conversation. The MCP servers folder contains two bundled integrations: one that gathers several Python packages for retrieving biological and biomedical data, and another that provides chemistry tools for editing molecular structures. The README treats this repository as configuration and runtime infrastructure rather than a finished product, and it includes guidance for anyone editing it, such as reading a skill's full instructions before changing it, preserving existing safety and validation rules, and keeping documentation in sync with any helper code a skill ships alongside its instructions.
A bundle of skills, agent profiles, and MCP server connectors that power Claude Science, an AI assistant for scientific computing.
Mainly HTML. The stack also includes Python, MCP, YAML.
Individual skills specify their own license, such as Apache-2.0, in their metadata.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.