Details
- Google AI Developers announced that Google DeepMind's Science Skills toolkit is now available as an open-source project on GitHub.
- Science Skills is a curated collection of agent skills for scientific research tasks, including genomics, structural biology, cheminformatics, and literature search.
- Each skill bundles structured instructions, scripts, and references so AI agents can reliably execute specialized scientific workflows rather than just answer questions.
- The toolkit is designed to improve scientific grounding and token efficiency for autonomous and semi-autonomous research agents.
- Developers can install the Science Skills bundle using npx skills add google-deepmind/science-skills and integrate it into compatible agent frameworks.
- Science Skills can be used within Google's Antigravity environment, where users can enable a Science plugin to access a curated subset of these skills for research workflows.
- Some skills integrate external scientific resources such as AlphaGenome, AFDB, UniProt, and other databases, with optional API keys for higher rate limits or full functionality.
- The accompanying demo and examples show how agents can chain these skills to handle end-to-end tasks like querying databases, analyzing structures, and synthesizing literature findings.
Impact
By open-sourcing Science Skills, Google DeepMind extends its scientific AI work into the broader agent ecosystem, pushing agentic workflows beyond generic chat toward reproducible, tool-aware scientific procedures. This move directly competes with emerging third-party scientific skill libraries and should make it easier for labs and startups to build AI scientists that plug into established bioinformatics and chemistry toolchains.
