How to Integrate google/skills into Your Application: A Complete Guide

Integrate google/skills into your application by installing via npx skills, selecting the specific skill modules you need, and registering them with your AI agent harness (Claude, Codex, or Antigravity).

The google/skills repository provides a collection of Agent Skills—self-contained, markdown-driven modules that extend AI agents with Google Cloud expertise. This guide walks you through the exact integration pattern used in the source code, from installation to runtime activation.

Install the Skills Package

The fastest way to bootstrap the repository is through the universal npx skills installer. This command resolves the dependency graph and places skill files where your agent runtime expects them.

# Pull the complete skills repository

npx skills add google/skills

This creates a local skills/ directory containing every skill defined in the repo. The installer is documented in README.md lines 13–16.

Install Only the Skills You Need

For production deployments, cherry-pick specific modules to minimize footprint:

# List all available skills

npx skills list

# Add targeted skills only

npx skills add google/skills --skill "cloud/cloud-run-basics" "cloud/bigquery-basics"

Each skill lives at skills/<category>/<skill-name>/SKILL.md. For example, the authentication helper skill resides at skills/cloud/google-cloud-recipe-auth/SKILL.md.

Select and Inspect Skills

Every skill is a markdown-based contract containing metadata, prompts, and reference links. Load these programmatically to extract prompt sections for your LLM pipeline.

import json
from pathlib import Path

# Load a skill's prompt definition

skill_path = Path("skills/cloud/google-cloud-recipe-auth/SKILL.md")
skill_content = skill_path.read_text(encoding="utf-8")

# Extract the "Examples" section to send to your LLM endpoint

examples_start = skill_content.find("## Examples")

examples = skill_content[examples_start:]

# Send the prompt to your LLM (pseudo-code)

response = llm_client.generate(
    prompt=examples,
    temperature=0.2,
)
print(json.dumps(response, indent=2))

Key skills referenced in the source analysis include:

• google-cloud-recipe-auth — Authentication flows and credential management
• gcloud — CLI-style wrappers for common gcloud actions
• bigquery-basics — BigQuery query patterns and code snippets
• cloud-monitoring-chart-generation — Programmatic monitoring dashboard creation

Activate Skills in Your Agent Harness

Registration varies by agent. The repository ships plugin definitions in the plugins/ directory for three major harnesses.

Claude Code Integration

# Add the repository to Claude's plugin marketplace

claude plugin marketplace add google/skills

# Install a specific skill package

claude plugin install cloud-monitoring-chart-generation@google-plugins

See README.md lines 162–166 for the exact marketplace commands.

Codex Integration

codex plugin marketplace add google/skills

After marketplace registration, install individual skills through the Codex UI.

Antigravity Integration

# Direct URL-based installation

agy plugin install https://github.com/google/skills/skills/cloud/gcloud

Runtime Skill Invocation

Once registered, agents invoke skills through their conversational API. Most skills expose a "Clarifying Questions" section that prompts context-gathering before delivering solutions.

For CLI-oriented skills, helper scripts are available directly:

# After installation, invoke bundled scripts

./skills/cloud/gcloud/gcloud.sh auth login

Key Files and Structure

Understanding the repository layout helps you navigate the skill ecosystem:

Path	Purpose
README.md	Installation instructions and plugin compatibility tables
skills/cloud/google-cloud-recipe-auth/SKILL.md	Authentication flows with usage checklists
skills/cloud/gcloud/SKILL.md	gcloud CLI wrappers for agent invocation
skills/cloud/bigquery-basics/SKILL.md	BigQuery query patterns and examples
skills/cloud/cloud-monitoring-chart-generation/SKILL.md	Monitoring chart generation workflows
plugins/	Claude, Codex, and Antigravity plugin manifests

Summary

• Install via npx skills add google/skills to pull the full repository or use --skill flags for selective installation
• Select skills from skills/<category>/<name>/SKILL.md based on your Google Cloud use case
• Register with your agent harness using marketplace commands (Claude/Codex) or direct URL installation (Antigravity)
• Invoke through conversational APIs or bundled CLI scripts; leverage "Clarifying Questions" for context-aware responses

Frequently Asked Questions

What is the google/skills repository?

The google/skills repository is a collection of Agent Skills—self-contained markdown modules that extend AI agents with structured Google Cloud knowledge. Each skill includes prompts, examples, and reference materials that agents can consume through standardized interfaces.

Can I use google/skills without an AI agent harness?

Yes. While designed for agent integration, you can programmatically load skill files as shown in the Python example above. Extract prompt sections from SKILL.md files and pass them directly to any LLM API.

How do I update skills after initial installation?

Re-run npx skills add google/skills to pull the latest versions. The installer resolves the dependency graph and updates modified skills in place. For selective updates, specify individual skills with the --skill flag.

Which AI agents support google/skills natively?

The repository provides native plugin support for Claude Code, Codex, and Antigravity. The plugins/ directory contains manifests for each harness. Other agents can consume skills by parsing the markdown files directly.

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