科学智能代理

⚙️ Prerequisites

• Python: 3.13+ for repository tooling; individual skill dependencies may support broader Python ranges
• uv: Python package manager (required for installing skill dependencies)
• Client: Any agent that supports the Agent Skills standard (Cursor, Claude Code, Gemini CLI, Codex, Google Antigravity, etc.)
• System: macOS, Linux, or Windows with WSL2
• Dependencies: Automatically handled by individual skills (check SKILL.md files for specific requirements)

🎯 Getting Started

Browse and install interactively

gh skill install K-Dense-AI/scientific-agent-skills

Install a specific skill directly

gh skill install K-Dense-AI/scientific-agent-skills scanpy

Update all installed skills

gh skill update --all ```

Installing uv

The skills use uv as the package manager for installing Python dependencies. Install it using the instructions for your operating system:

macOS and Linux:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Alternative (via pip):

pip install uv

After installation, verify it works by running:

uv --version

For more installation options and details, visit the official uv documentation.

---

Installation & Setup

Q: Do I need all the Python packages installed? A: No! Only install the packages you need. Each skill specifies its requirements in its SKILL.md file.

Q: What if a skill doesn't work? A: First check the Troubleshooting section. If the issue persists, file an issue on GitHub with detailed reproduction steps.

Q: Do the skills work offline? A: Database skills require internet access to query APIs. Package skills work offline once Python dependencies are installed.

💡 Quick Examples

Once you've installed the skills, you can ask your AI agent to execute complex multi-step scientific workflows. Here are some example prompts:

🔬 Use Cases

Contribution Guidelines

✅ Adhere to the Agent Skills Specification — Every skill must follow the official spec (valid SKILL.md frontmatter, naming conventions, directory structure) ✅ Include a quoted metadata.version value in every SKILL.md ✅ Increment metadata.version when updating an existing skill ✅ Maintain consistency with existing skill documentation format ✅ Ensure all code examples are tested and functional ✅ Follow scientific best practices in examples and workflows ✅ Update relevant documentation when adding new capabilities ✅ Provide clear comments and docstrings in code ✅ Include references to official documentation

Option 1: npx (all platforms)

Install Scientific Agent Skills with a single command:

npx skills add K-Dense-AI/scientific-agent-skills

This is the official standard approach for installing Agent Skills across all platforms, including Claude Code, Claude Cowork, Codex, Gemini CLI, Google Antigravity, Cursor, OpenClaw, NVIDIA NemoClaw, Hermes, Pi, and any other agent that supports the open Agent Skills standard.

Option 2: GitHub CLI (`gh skill`)

If you use the GitHub CLI (v2.90.0+), you can install skills with gh skill:

```bash

🔧 **Easy Integration**

• Simple Setup - Copy skills to your skills directory and start working
• Automatic Discovery - Your agent automatically finds and uses relevant skills
• Well Documented - Each skill includes examples, use cases, and best practices

🧪 Drug Discovery Pipeline

Goal: Find novel EGFR inhibitors for lung cancer treatment

Prompt:

Use available skills you have access to whenever possible. Query ChEMBL for EGFR inhibitors (IC50 < 50nM), analyze structure-activity relationships 
with RDKit, generate improved analogs with datamol, perform virtual screening with DiffDock 
against AlphaFold EGFR structure, search PubMed for resistance mechanisms, check COSMIC for 
mutations, and create visualizations and a comprehensive report.

Skills Used: database-lookup, rdkit, datamol, diffdock, paper-lookup, scientific-visualization

---

🔬 Single-Cell RNA-seq Analysis

Goal: Comprehensive analysis of 10X Genomics data with public data integration

Prompt:

Use available skills you have access to whenever possible. Load 10X dataset with Scanpy, perform QC and doublet removal, integrate with Cellxgene 
Census data, identify cell types using NCBI Gene markers, run differential expression with 
PyDESeq2, infer gene regulatory networks with Arboreto, enrich pathways via Reactome/KEGG, 
and identify therapeutic targets with Open Targets.

Skills Used: scanpy, cellxgene-census, database-lookup, pydeseq2, arboreto

---

🔧 Troubleshooting

❓ FAQ

General Questions

Q: Is this free to use? A: Yes! This repository is MIT licensed. However, each individual skill has its own license specified in the license metadata field within its SKILL.md file—be sure to review and comply with those terms.

Q: Why are all skills grouped together instead of separate packages? A: We believe good science in the age of AI is inherently interdisciplinary. Bundling all skills together makes it trivial for you (and your agent) to bridge across fields—e.g., combining genomics, cheminformatics, clinical data, and machine learning in one workflow—without worrying about which individual skills to install or wire together.

Q: Can I use this for commercial projects? A: The repository itself is MIT licensed, which allows commercial use. However, individual skills may have different licenses—check the license field in each skill's SKILL.md file to ensure compliance with your intended use.

Q: Do all skills have the same license? A: No. Each skill has its own license specified in the license metadata field within its SKILL.md file. These licenses may differ from the repository's MIT License. Users are responsible for reviewing and adhering to the license terms of each individual skill they use.

Q: How often is this updated? A: We regularly update skills to reflect the latest versions of packages and APIs. Major updates are announced in release notes.

Q: Can I use this with other AI models? A: The skills follow the open Agent Skills standard and work with any compatible agent, including Cursor, Claude Code, Codex, Google Antigravity, OpenClaw, NVIDIA NemoClaw, Hermes, and Pi.