This collection of agent skills enhances AI-driven scientific research by providing specialized tools and structured instructions for tasks across genomics, structural biology, cheminformatics, and literature search. It aims to improve AI agent grounding and token efficiency, enabling faster and more effective scientific workflows by integrating insights from over 30 databases and tools. The target audience includes AI researchers and developers building agentic systems for scientific discovery.

How It Works

Each skill is organized into a directory containing a main instruction file (SKILL.md), helper scripts, and optional references. This structure allows AI agents to access and execute specialized scientific tasks, leveraging curated resources and structured guidance. The approach focuses on providing deep grounding in scientific domains, enabling agents to interact with complex data sources and tools more effectively and with higher token efficiency.

Quick Start & Requirements

• Primary install: npx skills add google-deepmind/science-skills/
• Prerequisites: Requires the uv package manager (automatically installed on first skill trigger). Some skills (e.g., AlphaGenome, OpenAlex) require API keys for full functionality; others (e.g., ClinVar) benefit from API keys for higher rate limits. Users are guided to configure API keys, typically via a ~/.env file. Integration with Google Antigravity is also described.
• Links: Use cases and demo available at antigravity.google/use-cases/science.

Highlighted Details

• Integrates insights from AlphaGenome, AFDB, UniProt, and over 30 other scientific databases and tools.
• Provides structured instructions, scripts, and resources to extend AI agent capabilities for specialized scientific tasks.
• Aims to improve grounding and token efficiency in agentic scientific workflows.

Maintenance & Community

No specific details regarding community channels (e.g., Discord, Slack), roadmap, or notable contributors/sponsorships are provided in the README. The project is explicitly stated as "not an official Google product."

Licensing & Compatibility

• License: Software components are licensed under the Apache License, Version 2.0 (Apache 2.0). Other materials are licensed under the Creative Commons Attribution 4.0 International License (CC-BY).
• Compatibility: Apache 2.0 is permissive for commercial use and linking. Users must comply with the terms of both licenses and are responsible for adhering to the licenses of any third-party data sources referenced within individual skills.

Limitations & Caveats

The skills are designed for integration with AI agent frameworks like Google Antigravity and may require specific API key configurations for certain functionalities. Users bear responsibility for ensuring compliance with the licenses of all integrated third-party data sources. The project is identified as "not an official Google product."