google-deepmind

google-deepmind / science-skills

Public

GDM Science Skills to speed up agentic scientific workflows with better grounding and higher token efficiency. Integrate insights from AlphaGenome, AFDB, UniProt and 30+ other databases and tools.

antigravity.googleuse-casesscience
14
2
100% credibility
Found May 19, 2026 at 23 stars 2x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Google DeepMind Science Skills is a curated collection of specialized research tools designed to extend an AI assistant's capabilities for scientific investigation. The collection includes over 30 different skills covering genomics, structural biology, cheminformatics, literature search, and clinical data. Each skill provides structured instructions and helper scripts that allow an AI agent to query trusted scientific databases (like AlphaFold, ClinVar, PubMed, ChEMBL, and many others), retrieve relevant data, perform analyses, and present results in formats suitable for research. The project is designed to work with Google Antigravity and is licensed under Apache 2.0 with clear terms of use for third-party data sources.

How It Works

1
🔬 You have a scientific question

You need to analyze a protein structure, look up gene variants, search research papers, or explore drug interactions for your research.

2
🤖 You ask your AI research assistant

You describe your research question naturally, like asking a knowledgeable colleague for help with your scientific investigation.

3
🧬 Your assistant springs into action

Behind the scenes, your AI connects to trusted scientific databases—protein structures, gene databases, medical records, research papers—and gathers exactly what you need.

4
Your assistant prepares your results
🧪
Protein & Structure Analysis

Fetch and analyze protein structures, find similar proteins, identify structural domains and flexibility.

🧬
Genomics & Variants

Look up gene information, check variant frequencies in populations, understand clinical significance of mutations.

📚
Literature Research

Search through millions of research papers, find relevant studies, download full articles.

💊
Drug & Compound Data

Explore drug interactions, find similar molecules, check FDA safety reports.

5
✨ You receive clear, organized results

Your assistant presents findings in easy-to-understand formats—tables, charts, summaries—along with links to original sources so you can verify everything.

🎯 You continue your research with confidence

You now have verified scientific data, properly cited and formatted, ready to use in your research paper, presentation, or next discovery.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 23 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is science-skills?

GDM Science Skills is a collection of AI agent tools for scientific research, spanning genomics, structural biology, cheminformatics, and literature search. It wraps APIs from AlphaGenome, AlphaFold, UniProt, ClinVar, dbSNP, ChEMBL, PubMed, and 30+ other scientific databases into ready-to-use skill modules. Each skill provides structured instructions and helper scripts that let an AI agent query specialized scientific data without manual lookup. Built in Python, it uses the `uv` package manager and is designed to integrate with Google Antigravity, though the scripts can run standalone from the command line.

Why is it gaining traction?

The pitch is compelling: instead of manually hunting through genomic databases or wrestling with API documentation, you delegate the work to an AI agent that can fetch protein structures, look up gene annotations, search clinical trial databases, and analyze variant effects through a unified interface. The skills handle the plumbing between large models like AlphaGenome and the messy reality of scientific data sources, normalizing units, handling rate limits, and parsing complex API responses. It addresses a real pain point for computational biologists and bioinformaticians who spend too much time on data retrieval rather than analysis.

Who should use this?

This is squarely aimed at researchers in computational biology, genomics, and drug discovery who want to prototype AI-assisted workflows. A bioinformatician building a variant annotation pipeline would find the ClinVar and dbSNP integration useful. A drug researcher could leverage the ChEMBL and PubChem wrappers to pull bioactivity data. That said, with only 14 stars and a 1.0% credibility score, this is early-stage software. The documentation exists, but test coverage and battle-testing in production environments remain unknown.

Verdict

If you're experimenting with agentic scientific workflows, this is worth a careful look, especially for genomics and protein structure tasks. But approach it as exploration material rather than production infrastructure. The 1.0% credibility score signals that community validation is minimal, and the low star count suggests this is either very new or still finding its audience. Review the Apache 2.0 license and verify API key requirements for AlphaGenome before committing.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.