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.
2.3k
Stars
234
Forks
1
Open issues
3
Contributors
AI Analysis
Science Skills is a curated collection of agent skills for scientific research tasks spanning genomics, structural biology, cheminformatics, and literature search, designed to extend AI agent capabilities for specialized scientific workflows. It integrates insights from 30+ databases including AlphaGenome, AFDB, and UniProt, and is specifically built for Google Antigravity and compatible AI agents. This project is narrowly focused on scientific researchers and AI agents performing scientific ...
Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.
AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.
Google DeepMind's curated skill library for AI agents in scientific research, integrated with Antigravity platform
Science Skills is a collection of structured agent skills for genomics, structural biology, cheminformatics, and literature search, designed to ground AI agents in scientific workflows. Built by Google DeepMind and tightly integrated with Google Antigravity (a proprietary agentic platform), it provides pre-packaged access to 30+ scientific databases (AlphaGenome, UniProt, AFDB, ClinVar, OpenAlex). Adoption appears to be coupled to Antigravity uptake; real-world usage outside that ecosystem is unverified.
Launched 2026-05-13, very recent. Appears to be part of a broader Google push into agentic scientific workflows, following AlphaGenome (2025) and other agent-infrastructure projects. Positioned as a skills layer for Antigravity rather than a standalone library.
2,098 stars and 153 new stars in 7 days indicates rapid early traction, but this is within 6 weeks of creation. Growth likely driven by Google DeepMind brand, integration with Antigravity platform announcement, and timing within broader agent-skills ecosystem momentum. Lack of historical data makes trend prediction unreliable.
Adoption not verified. README describes integration with Google Antigravity (proprietary platform) and references use cases at antigravity.google/use-cases/science, but no independent evidence of real-world scientific workflows using these skills, publication rates, or user testimonials. Some skills require API keys (AlphaGenome, OpenAlex), suggesting intent for production use, but actual deployment scale is undocumented.
Based on README, appears to be a modular skill collection where each skill is a directory containing SKILL.md (YAML frontmatter + markdown instructions), scripts/, and optional references/. Skills are installed via npx/skills.sh tooling and executed within Antigravity or compatible agent runtimes. Likely uses Python with `uv` package manager for dependency isolation. Structured as skill definitions rather than a monolithic library.
Not documented in README. No mention of testing framework, CI/CD validation, or test coverage metrics.
Last push 2026-06-08 (20 days before analysis date) indicates recent activity. Repository is less than 2 months old, so maintenance patterns are not yet established. Apache 2.0 licensing and Google DeepMind authorship suggest institutional backing, but repo maturity cannot be assessed from age alone. Frequency of push activity cannot be evaluated from single creation and last-push dates.
ADOPT IF: you are already a Google Antigravity user needing to accelerate scientific research workflows with agent skills, and the curated skill set (genomics, structural biology, cheminformatics) aligns with your domain. AVOID IF: you need to operate independently of Google's proprietary platform, require battle-tested production infrastructure with transparent, multi-year operational history, or need a framework that works with open-source or competing agent runtimes. MONITOR IF: you are evaluating agent infrastructure for scientific research; this will likely become the default science-domain skill layer if Antigravity adoption grows, but platform lock-in and lack of open alternatives outside Antigravity are material concerns.
Independent dimensions
Mainstream potential
4/10
Technical importance
6/10
Adoption evidence
2/10
- Platform dependency: tightly coupled to Google Antigravity; unclear if skills work with other agent frameworks or if they can be used standalone.
- Adoption not verified: no public evidence of production deployment at scale; growth metrics reflect early GitHub traction rather than real-world usage.
- Rapid age and incomplete maturity: repository less than 2 months old; maintenance patterns, breaking-change policy, and long-term support commitment are unproven.
- API key requirements and third-party rate limits: many skills depend on external services (AlphaGenome, OpenAlex, ClinVar) with their own quotas and licensing terms; cascading failures possible if external APIs become unavailable or restrictive.
- Vendor lock-in and licensing complexity: SKILL_LICENSES.md references multiple third-party data source terms; compliance responsibility on user; Google DeepMind project status as 'not an official Google product' may affect support guarantees.
Science Skills will likely deepen integration with Antigravity and expand to 40–50+ scientific skills within 12 months if Antigravity adoption accelerates. Star count may plateau or decline if adoption remains confined to proprietary platform. Key dependency: Antigravity's own market traction and whether Google commits long-term resources to the agentic platform.
Newsletter
Get analyses like this every Monday
Free weekly digest of the most interesting open-source discoveries.
Languages
Information
- Language
- Python
- License
- Apache-2.0
- Last updated
- 3d ago
- Created
- 2mo ago
- Analyzed with
- anthropic/claude-haiku-4-5
Stars over time
Contributors over time
Top 100 contributors only — repos with more will plateau at 100.
Top contributors
Similar repos
K-Dense-AI/scientific-agent-skills
Scientific Agent Skills is a collection of 148 ready-to-use skills and 100+...
aipoch/medical-research-skills
A curated library of 550+ agent skills for medical research, enabling AI...
KKKKhazix/khazix-skills
A collection of structured AI Skills for agent-based workflows (Claude Code,...
| Repository | Stars | Week Δ | Language | Score | Updated |
|---|---|---|---|---|---|
|
|
2.3k | +138 | Python | 7/10 | 3d ago |
|
|
30.6k | — | Python | 8/10 | 2d ago |
|
|
1.3k | — | Python | 8/10 | 6h ago |
|
|
16.8k | — | Python | 7/10 | 1w ago |
|
|
14.5k | — | Python | 7/10 | 14h ago |
|
|
2k | — | Python | 8/10 | 21h ago |
More established and widely starred, but appears to be a more general agent-skills framework. Science Skills is domain-focused on scientific research with deeper integration to specific databases (AlphaGenome, UniProt, AFDB). Different positioning: K-Dense broader framework vs. Science Skills curated domain layer.
Likely the parent or sibling framework for agent skills. Science Skills appears to be a specialized, curated subset for scientific tasks, not a direct replacement.
AlphaGenome is a specific capability (protein structure prediction); Science Skills integrates it as one of 30+ skills. Science Skills is a consumption/coordination layer, not a replacement.
Narrower domain (medical research). Science Skills broader (genomics, cheminformatics, structural biology, literature). Science Skills backed by Google DeepMind; adoption trajectory and independence of aipoch project unclear.