Google Agent Skills Reference

Key takeaways#

• google/skills is Google’s public Agent Skills repository for Google products and technologies.
• The repository is a skills collection, not a standalone AI app or model.
• Installation is shown with npx skills add google/skills, then users can select individual skills.
• The catalog includes Google Cloud onboarding, GKE, BigQuery, AlloyDB, Agent Platform, Gemini, RAG, monitoring, logging, security, and cost optimization skills.
• Use it as a source-backed reference for agent operators building repeatable Google Cloud and Gemini workflows.

What this resource is#

Google Agent Skills is a public repository of Agent Skills for Google products and technologies. The README describes it as a collection of skills for Google Cloud and related agent workflows, with installation through the skills.sh ecosystem. It should be modeled as a resource because the durable entity is a catalog of reusable agent instructions, recipes, and operational playbooks. It is not a hosted SaaS product, not an MCP server, and not a model family.

The repository is useful for builders who want agents to follow specific Google Cloud procedures instead of relying on generic model memory. Skills give an agent more structured instructions for tasks such as onboarding to Google Cloud, building and deploying AI agents, working with GKE inference, using BigQuery AI and ML, configuring Cloud Logging, tuning agent platform models, setting up RAG workflows, and handling operational troubleshooting.

How to install#

The README shows this install command:

npx skills add google/skills

After running the command, the user can select specific skills from the repository. That matters because the collection is large. Installing only the relevant skills keeps an agent’s working context focused and reduces the chance that unrelated operational guidance is pulled into a task.

What is included#

The visible catalog includes multiple groups of skills. Cloud setup and onboarding skills cover Google Cloud authentication, onboarding, foundation builder workflows, architecture patterns, networking, storage, and serverless application tiers. AI-specific skills cover building and deploying AI agents on Google Cloud, data science workflows with AI agents, live bidirectional multimodal streaming, migration from Gemini API in AI Studio to Agent Platform, model garden deployment, prompt management, RAG engine management, tuning, evaluations, and troubleshooting.

There are also specialized operational skills for GKE, BigQuery, AlloyDB, Bigtable, Cloud SQL, Spanner, Cloud Monitoring, Cloud Logging, Workload Manager, SecOps detection coverage, cost analysis, and Well-Architected Framework pillars. For builders, the important point is not the raw count. The useful point is that the repository turns repeatable Google platform procedures into agent-readable skill packages.

When to use it#

Use Google Agent Skills when an agent needs repeatable guidance for Google Cloud, Gemini, Agent Platform, GKE, BigQuery, RAG, logging, monitoring, security, or cost work. It is especially useful for teams that already use AI coding agents and want safer, more specific instructions for cloud operations. The skills can help standardize prompts and procedures across people and agent sessions.

Do not treat the repository as a substitute for official product documentation, IAM review, or production change approval. Skills can guide an agent, but humans still need to review cloud permissions, billable resources, production changes, data access, and security boundaries. For sensitive environments, start in a sandbox project and inspect the skill text before relying on it.

Verification notes#

This resource was created from the official public repository at https://github.com/google/skills. Repository metadata changes quickly, so users should confirm the README, install command, license, skill list, and any referenced product docs before using the skills in production agent workflows.