academic-skills
深度云创科技，专注开发 AI 智能体、AI 应用辅助科研、工作、学习！
Collection of 139 installable skill directories that teach AI coding agents (Claude Code, Codex, OpenCode) how to use scientific tools and research workflows across bioinformatics, drug discovery, materials science, and more.
This repository is a collection of 139 packaged skill directories for AI coding agents, focused on scientific and academic work. Each skill teaches an AI agent how to use a specific tool, library, or research workflow. The collection covers a wide range of fields: bioinformatics and genomics, drug discovery and molecular modeling, clinical research, geospatial science, materials science, physics, astronomy, and more. It also includes skills for tasks like literature review, scientific writing, citation management, and AI model interpretation.
Each skill directory holds a SKILL.md file that describes what the agent should do, along with optional scripts, reference files, example data, or tests. The skill format is designed to be portable across different AI coding agents. The repository explicitly supports three agents: Claude Code (Anthropic), Codex (OpenAI), and OpenCode. Installation is handled via a command-line tool called npx skills, which copies the selected skill directories into the agent's local configuration folder. You can install all 139 skills at once or pick individual ones by name.
The install paths differ by agent. Claude Code reads skills from .claude/skills/ inside your project or from ~/.claude/skills/ for a user-level install. Codex uses .agents/skills/ for projects. OpenCode has its own native paths but also reads from the Claude Code and shared agent paths. After installing, a Python validation script checks compatibility so you can confirm the skills landed correctly before using them.
The README includes a security caution worth noting: skills can instruct an AI agent to run shell commands, install packages, make network requests, read environment variables, or modify files. The repository includes a pre-run security scan of its own skill directories, and you can run the Cisco AI skill scanner yourself to audit the content before installing anything.
This is a fork of an upstream open-source skills collection, cleaned up to remove organization-specific install paths and tracking. It carries an MIT license, and the original copyright notice is preserved as the license terms require.
Where it fits
- Install all 139 research skills into Claude Code to immediately give your AI assistant working knowledge of bioinformatics, molecular modeling, and other scientific tools
- Pick individual skills by name to extend your agent with a specific workflow like literature review, citation management, or a genomics pipeline
- Audit the skill content with the bundled Cisco AI scanner before installing in a sensitive research environment to check for unintended shell commands or network calls