How to Integrate google/skills with Other Google Cloud Services: A Complete Guide

Integrating google/skills with Google Cloud services requires establishing authentication via Application Default Credentials, selecting a modular skill from the repository, and executing service-specific workflows through gcloud commands or Python client libraries.

The google/skills repository provides a collection of modular, Markdown-driven skills—self-contained recipes that guide AI agents (or operators) through complete Google Cloud operations. Each skill demonstrates practical integration patterns with core services like BigQuery, Cloud Storage, Vertex AI, and Cloud Monitoring. This guide walks you through the architecture, authentication flows, and concrete code examples needed to connect these skills to your existing Google Cloud infrastructure.

---

Authentication and Identity Foundation

Every skill integration begins with proper authentication. The repository centralizes credential handling in the Authenticating to Google Cloud skill located at skills/cloud/google-cloud-recipe-auth/SKILL.md.

Two primary authentication patterns are supported:

• Application Default Credentials (ADC) – The standard flow for local development, GCE VMs, GKE Workloads, and Cloud Run services. Run gcloud auth application-default login to populate credentials locally.

• Service-Account Impersonation & Workload Identity Federation – For production workloads and external identity providers. This enables secure, short-lived tokens without exporting service account keys.

The authentication skill also implements lazy IAM remediation—automatic permission fixes when a call fails due to missing roles.

# Authenticate with ADC (Python)

from google.auth import default
from google.auth.transport.requests import Request

creds, project = default()
creds.refresh(Request())
print(f"Authenticated to project: {project}")

This snippet works identically across laptops, Cloud Shell, GKE, and Cloud Run environments.

---

Service-to-Service Integration Patterns

Once authenticated, skills invoke Google Cloud APIs through two primary mechanisms: direct client library calls and gcloud CLI orchestration.

Direct API Calls with Client Libraries

The Agent-Platform Inference folder contains production-ready Python wrappers for Vertex AI. The file skills/cloud/agent-platform-inference/scripts/gemini_vertexai_sdk.py demonstrates how to call Gemini models and forward results to downstream services.

# Generate Gemini response and store in BigQuery

import json
from google.cloud import bigquery
from skills.cloud.agent_platform_inference.scripts.gemini_vertexai_sdk import (
    generate_text,
)

# Call Vertex AI Gemini

prompt = "Explain the benefits of Workload Identity Federation."
response = generate_text(prompt)

# Persist to BigQuery

client = bigquery.Client()
table_id = "my-project.my_dataset.gemini_responses"
rows_to_insert = [
    {
        "timestamp": bigquery.RowIterator()._client._datetime.utcnow().isoformat(),
        "prompt": prompt,
        "response": response,
    }
]

errors = client.insert_rows_json(table_id, rows_to_insert)
if errors:
    raise RuntimeError(f"BigQuery insert errors: {errors}")

This pattern—model invocation → structured storage → error handling—is reusable across any generative AI workflow.

gcloud-Based Resource Provisioning

The Foundation Builder skill (skills/cloud/google-cloud-recipe-foundation-builder/SKILL.md) demonstrates cross-service orchestration. It provisions organization-level resources, links projects to billing accounts, enables APIs, and configures centralized logging.

# Capture organization and billing context

ORG_ID=$(gcloud organizations list --format="value(ID)" --limit=1)
BILLING_ID=$(gcloud billing accounts list --filter=open=true \
               --format="value(ACCOUNT_ID)" --limit=1)

# Install and execute the skill

npx skills add google/skills

# Select "google-cloud-recipe-foundation-builder"

# Provide: Organization ID, Billing Account ID, project suffix, log region

The skill automatically creates folder hierarchies, projects, enables required APIs, and sets up log sinks—demonstrating how a single skill coordinates multiple Google Cloud services.

---

MCP: Model Control Plane Integration

Many skills expose MCP (Model Control Plane) endpoints that wrap underlying Google Cloud service APIs. These provide reusable, CLI-style commands for common operations.

The BigQuery MCP, documented in skills/cloud/google-cloud-networking-observability/references/mcp-usage.md, offers helpers like:

• list_dataset_ids – Enumerate datasets in a project
• list_table_ids – List tables within a dataset

MCP utilities reduce boilerplate by abstracting discovery and pagination logic. A skill can import these helpers rather than reimplementing BigQuery client calls from scratch.

---

Observability and Monitoring Integration

Skills can generate Cloud Monitoring dashboards and export telemetry to BigQuery for downstream analytics. The Cloud-Monitoring-Chart-Generation scripts in skills/cloud/cloud-monitoring-chart-generation/scripts/assemble_widget_proto.py assemble protobuf widgets programmatically.

Key capabilities include:

• Building time-series charts, scorecards, and alert policies
• Optional persistence of metric data to BigQuery datasets
• Integration with existing monitoring workspaces

This enables skills to not only perform operations but also create persistent observability artifacts for long-running infrastructure.

---

CI/CD and Infrastructure Automation

The Design-Deploy skill (skills/cloud/design-deploy/references/design/SKILL.md) provides Terraform-based templates for infrastructure-as-code workflows. These integrate with:

• Cloud Build – Native Google Cloud CI/CD
• GitHub Actions – External pipeline integration
• Custom pipeline tools – Via shell script invocations

The template-driven approach allows skills to generate reproducible infrastructure definitions that slot into existing deployment pipelines.

---

Step-by-Step Integration Workflow

Follow this sequence when integrating google/skills with Google Cloud services:

Step	Action	Key File
1. Authenticate	Configure ADC or Workload Identity Federation	skills/cloud/google-cloud-recipe-auth/SKILL.md
2. Select Skill	Use npx skills add google/skills to browse available skills	README.md
3. Execute Skill	Follow interactive prompts in the skill's SKILL.md	Skill-specific SKILL.md
4. Invoke Service	Use client libraries or gcloud commands as orchestrated by the skill	skills/cloud/agent-platform-inference/scripts/gemini_vertexai_sdk.py
5. Persist Results	Store outputs in BigQuery, Cloud Storage, or Monitoring	skills/cloud/cloud-monitoring-chart-generation/scripts/assemble_widget_proto.py
6. Handle Errors	Leverage built-in IAM remediation for permission failures	skills/cloud/google-cloud-recipe-foundation-builder/SKILL.md

---

Summary

• google/skills provides modular, Markdown-based skills for Google Cloud operations—each skill is a self-contained integration guide with executable code.

• Authentication is mandatory and standardized through the Authenticating to Google Cloud skill, supporting ADC, service account impersonation, and Workload Identity Federation.

• Integration patterns span three layers: direct client library calls (Python SDKs), gcloud CLI orchestration, and MCP helper utilities for service abstraction.

• Cross-service workflows are demonstrated in Foundation Builder (resource provisioning), Agent-Platform Inference (AI model calls), and Cloud-Monitoring-Chart-Generation (observability).

• CI/CD integration via Design-Deploy skills enables infrastructure-as-code pipelines with Terraform templates compatible with Cloud Build and GitHub Actions.

---

Frequently Asked Questions

What is the fastest way to start integrating google/skills with my existing Google Cloud project?

Run gcloud auth application-default login to establish credentials, then execute npx skills add google/skills to browse and install skills. Start with the Foundation Builder skill to provision a properly configured project with billing, APIs, and logging enabled. This single skill demonstrates integration patterns applicable to any subsequent service-specific skill.

Can google/skills integrate with services outside Google Cloud?

Yes, through Workload Identity Federation. The authentication skill supports external identity providers (AWS, Azure AD, on-premises OIDC), allowing skills to obtain Google Cloud credentials without service account keys. Additionally, skills like Agent-Platform Inference include OpenAI-compatible endpoints alongside Vertex AI, enabling multi-provider AI workflows.

How do skills handle permission errors during Google Cloud API calls?

The Foundation Builder skill implements lazy IAM remediation—it detects PERMISSION_DENIED errors, analyzes required roles, and attempts automatic fixes when the executing identity has sufficient administrative privileges. This pattern is documented in phase 2 of the skill's SKILL.md and reduces manual IAM troubleshooting during automation workflows.

What is MCP in the context of google/skills?

MCP (Model Control Plane) refers to helper utilities that wrap Google Cloud service APIs into reusable command interfaces. The BigQuery MCP in skills/cloud/google-cloud-networking-observability/references/mcp-usage.md exposes functions like list_dataset_ids and list_table_ids that skills can import rather than implementing raw BigQuery client calls. This abstraction layer promotes code reuse across multiple skills.

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