How to Use the `google/skills` Repository: Installation, Access, and Sample Usage

The google/skills repository provides Markdown-based Agent Skills that power AI-driven assistants to answer questions, generate commands, and guide users through Google Cloud tasks.

The google/skills repository is a collection of Agent Skills—structured Markdown recipes that describe how to work with Google Cloud services, data platforms, and other Google products. These skills are not traditional code libraries. Instead, they serve as knowledge sources for AI agents such as Claude, Codex, and Antigravity, enabling them to answer queries, generate commands, and walk users through complex cloud workflows.

What Are Agent Skills?

Agent Skills in the google/skills repository follow a three-layer architecture:

• Skill Markdown Files – Each SKILL.md contains structured front-matter (name, metadata) followed by human-readable instructions, clarification questions, and concrete examples including Python snippets and gcloud commands.

• Reference Documents – Adjacent references/ folders hold supplemental material such as Terraform snippets, CLI usage guides, and security best practices that skills can reference.

• Agent Harness Integration – The repository bundles plugins that expose skills to various AI agents, making them discoverable through commands like npx skills list or skills add.

Installing and Accessing google/skills

To begin using google/skills, install the skill set through the skills CLI:

npx skills add google/skills

This command downloads the repository into your local skills cache and makes all SKILL.md files available to the agent harness.

Exploring Available Skills

After installation, list all available skills:

npx skills list

This displays a tree of all skills, mirroring the Available Skills section of the repository's README.md.

Registering with AI Agents

To use google/skills with Claude, register the repository as a plugin:

claude plugin marketplace add google/skills
claude plugin install <plugin>@google-plugins

According to the google/skills source code, plugins for Claude, Codex, and Antigravity can be registered via marketplace commands (see README.md lines 59-66).

Example Usage: Authentication Skill

One of the most fundamental skills is google-cloud-recipe-auth, located at skills/cloud/google-cloud-recipe-auth/SKILL.md. This skill explains authentication flows with runnable examples.

Python Example: List Cloud Storage Buckets with ADC

First, authenticate locally:

gcloud auth application-default login

Then run Python code that automatically discovers your credentials:

from google.cloud import storage

client = storage.Client()          # ADC automatically discovered

for bucket in client.list_buckets():
    print(bucket.name)

This example appears in the Human-to-Service (Local Python Development) section of the authentication skill (lines 9-16 of SKILL.md).

gcloud Example: Impersonate a Service Account

For scenarios requiring service account impersonation:

# Use your own user credentials to impersonate a service account

gcloud auth login
gcloud config set account your.email@example.com
gcloud auth activate-service-account \
  --impersonate-service-account=my-sa@my-project.iam.gserviceaccount.com

This workflow is outlined in the Human Authentication section (lines 70-78 of the same SKILL.md).

Infrastructure as Code Examples

Many skills include Terraform configurations in their references/ folders.

Spanner Instance Provisioning

From skills/cloud/spanner-basics/references/terraform-usage.md (lines 19-28):

resource "google_spanner_instance" "example" {
  name         = "example-instance"
  config       = "regional-us-central1"
  display_name = "Example Spanner Instance"
  node_count   = 1
}

resource "google_spanner_database" "example" {
  name          = "example-database"
  instance_name = google_spanner_instance.example.name
}

Apply this configuration with:

terraform init
terraform apply -var='project_id=my-project'

Kubernetes Configuration Example

The GKE workload scaling skill includes complete YAML manifests in its assets/ folder.

From skills/cloud/gke-workload-scaling/assets/hpa-example.yaml (lines 1-20):

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: example-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: example-deployment
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50

Key Files in google/skills

File Path	Purpose
README.md	Overview, install commands, and complete skill index
skills/cloud/google-cloud-recipe-auth/SKILL.md	Authentication/authorization guide with runnable Python and gcloud examples
skills/cloud/spanner-basics/references/terraform-usage.md	Terraform patterns for Spanner provisioning
skills/cloud/gke-workload-scaling/assets/hpa-example.yaml	Production-ready HorizontalPodAutoscaler manifest
skills/cloud/gke-upgrades/SKILL.md	GKE upgrade workflows with CLI commands and validation steps

Summary

• Installation: Use npx skills add google/skills to make all skills available locally.
• Discovery: Run npx skills list to browse available skills or consult the README.md Available Skills section.
• Structure: Each skill combines a front-matter SKILL.md with supplemental code in references/ or assets/ folders.
• Integration: Register plugins with your AI agent (Claude, Codex, Antigravity) to enable query-based access to skill content.
• Execution: Copy code examples directly from skill files—they are designed to run after completing any prerequisite authentication steps.

Frequently Asked Questions

What is the difference between google/skills and a Python package?

google/skills contains Markdown-based recipes that describe how to use Google Cloud services, not importable code. You install it via the skills CLI to make these recipes available to AI agents, then copy code snippets from the skill files into your own projects.

How do I find the right skill for my use case?

Run npx skills list after installation to see all available skills, or browse the Available Skills table in README.md. Skills are organized by service (Cloud, GKE, Spanner, etc.) and include descriptive front-matter naming each capability.

Can I use google/skills without an AI agent?

Yes. While designed for agent integration, the SKILL.md files and references/ folders contain standalone documentation and runnable code examples. You can read them directly and execute the included Python, gcloud, or Terraform snippets.

Where are the actual code examples stored?

Concrete examples live in three places: inline within SKILL.md files (especially in example sections), references/ folders (Terraform configurations and CLI guides), and assets/ folders (complete YAML manifests and configuration files).

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}