← all repositories
google/skills

Google's pre-packaged prompts for cloud LLM agents

A repo of reusable "skills" that teach agent frameworks how to talk to Google Cloud APIs and follow its well-architected pillars.

12k stars Python AgentsLLMOps · Eval
skills
Velocity · 7d
+175
★ / day
Trend
steady
star history

What it does This repository bundles structured prompt modules — “Agent Skills” — for Google products and technologies, mostly Google Cloud. Install them via npx skills add google/skills and pick the ones you need. Each skill is essentially a recipe that tells an LLM agent how to interact with a specific API or follow a specific operational practice.

The interesting bit Google is treating prompt engineering as infrastructure. Rather than hoping your agent figures out BigQuery or GKE on its own, you slot in a pre-built skill — like a driver for a very chatty, very forgetful coworker. The Well-Architected Framework skills (security, reliability, cost, etc.) are the unusual ones: they don’t teach APIs, they teach an agent to audit your architecture against Google’s own pillars.

Key highlights

  • Covers Gemini API, AlloyDB, BigQuery, Cloud Run, Cloud SQL, Firebase, GKE, and more
  • Includes “recipe” skills for onboarding, auth, and network observability
  • Six Well-Architected Framework skills for operational review
  • Installed via npx skills add, not pip or npm directly
  • Apache 2.0 licensed

Caveats

  • Repository is under active development; skill definitions may shift
  • Actual skill content lives in subdirectories; the README is just a catalog
  • It’s unclear which agent frameworks consume these skills or how compatibility works

Verdict Worth a look if you’re building agents that operate Google Cloud infrastructure and want to outsource the prompt-crafting. Skip it if you’re not using an agent framework that supports the skills.sh ecosystem — it’s not standalone tooling.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.