欢迎关注B站及YouTube频道：深度云创科技，感兴趣的朋友欢迎加入新时代智能体交流社群

客服微信：16773345788

Academic Skills

Academic Skills is a curated collection of agent skills for research, science, engineering, analysis, finance, writing, and scientific communication.

This repository is prepared as a neutral fork of an open-source Agent Skills collection. It keeps the reusable skill content and removes upstream organization-specific promotion, install paths, and integration tracking.

What Is Included

This snapshot contains 139 skill directories under academic-skills/.

The collection covers:

• Bioinformatics, genomics, single-cell analysis, and multi-omics workflows
• Cheminformatics, drug discovery, molecular modeling, and protein engineering
• Scientific Python packages such as RDKit, Scanpy, scikit-learn, PyMC, NetworkX, Matplotlib, SymPy, Dask, Astropy, and many others
• Scientific databases and lookup workflows
• Geospatial science, remote sensing, materials science, physics, and astronomy
• Clinical research, treatment planning, regulatory writing, and medical imaging
• Literature review, peer review, scientific writing, posters, slides, diagrams, citation management, and document processing
• AI, machine learning, forecasting, reinforcement learning, and model interpretation workflows

Each skill usually includes a SKILL.md file plus optional examples, scripts, references, assets, or tests.

Getting Started

These instructions install the portable Agent Skills layer: each skill directory contains a SKILL.md file plus optional scripts/, references/, assets/, or tests/ content. Agent-specific features such as Claude Code hooks, subagent forks, dynamic shell injection, Codex plugin packaging, and OpenCode permission policy are intentionally out of scope for this repository.

Check Compatibility

Validate the repository before installing:

python scripts/validate_compat.py

Preview available skills without installing:

npx skills add crazymsn/academic-skills --list --full-depth

Claude Code

Install every skill for Claude Code into the current project:

npx skills add crazymsn/academic-skills --skill '*' --agent claude-code --copy -y --full-depth

Install globally for the current user:

npx skills add crazymsn/academic-skills --skill '*' --agent claude-code --global --copy -y --full-depth

Claude Code project installs use .claude/skills/; user installs use ~/.claude/skills/ unless CLAUDE_CONFIG_DIR points elsewhere.

Verify a project install:

python scripts/validate_compat.py --agent claude --scope project

Codex

Install every skill for Codex into the current project:

npx skills add crazymsn/academic-skills --skill '*' --agent codex --copy -y --full-depth

Install globally for the current user:

npx skills add crazymsn/academic-skills --skill '*' --agent codex --global --copy -y --full-depth

Codex project installs use .agents/skills/. Depending on the Codex surface and installer version, user installs may be read from ~/.agents/skills/ or ~/.codex/skills/; the compatibility checker inspects both.

Verify a project install:

python scripts/validate_compat.py --agent codex --scope project

OpenCode

Install every skill for OpenCode into the current project:

npx skills add crazymsn/academic-skills --skill '*' --agent opencode --copy -y --full-depth

Install globally for the current user:

npx skills add crazymsn/academic-skills --skill '*' --agent opencode --global --copy -y --full-depth

OpenCode’s native project path is .opencode/skills/ and its native user path is ~/.config/opencode/skills/. It also reads compatible skills from .agents/skills/, .claude/skills/, ~/.agents/skills/, and ~/.claude/skills/; the npx skills installer currently targets the shared .agents/skills/ project path for OpenCode.

Verify a project install:

python scripts/validate_compat.py --agent opencode --scope project

Install Selected Skills

Install one skill:

npx skills add crazymsn/academic-skills --skill scanpy --agent codex --copy -y --full-depth

Install several skills:

npx skills add crazymsn/academic-skills --skill scanpy anndata scvi-tools --agent codex --copy -y --full-depth

Verify selected skills:

python scripts/validate_compat.py --skill scanpy --skill anndata --skill scvi-tools

Manual Install

You can also copy an individual directory from academic-skills/ into an agent’s local skills directory. For example, copy academic-skills/scanpy/ as one complete skill directory.

Security Notes

Skills can influence an AI coding agent to run commands, install packages, make network requests, read environment variables, or modify files. Review each skill before installing it, especially if it includes scripts, shell commands, third-party API calls, or unpinned dependency installation instructions.

A local review for this fork is summarized in SECURITY.md.

To run your own scan:

uv pip install cisco-ai-skill-scanner
skill-scanner scan academic-skills --use-behavioral

Repository Layout

academic-skills/      Skill directories
docs/                   Generated indexes and ecosystem notes
scripts/                Compatibility and install validation helpers
scan_skills.py          Security scan helper
scan_pr_skills.py       Pull-request scan helper
pyproject.toml          Python project metadata
uv.lock                 Locked Python dependency metadata

Maintenance Checklist

• Review new or changed SKILL.md files before merging.
• Prefer pinned dependency versions for executable examples.
• Avoid hardcoded credentials, tokens, account identifiers, and tracking headers.
• Keep generated security reports current after large skill updates.
• Preserve required third-party license notices.

License

This repository is distributed under the MIT License. See LICENSE.md.

The original MIT copyright notice is intentionally retained in the license file, as required when redistributing or modifying MIT-licensed software.