How to use the GCP Cloud Architect agent skill

What does the gcp-cloud-architect skill make the agent do?

The skill tells the agent to work through a fixed workflow: gather requirements, design an architecture, estimate cost, generate infrastructure-as-code, configure CI/CD, and run a security review. Its description covers Cloud Run, GKE, Cloud Functions, Cloud SQL, BigQuery and cost optimization.

It supports four recommended patterns: Serverless Web (Cloud Storage, Cloud CDN, Cloud Run, Firestore), Microservices on GKE (GKE Autopilot, Cloud SQL, Memorystore, Cloud Pub/Sub), Serverless Data Pipeline (Pub/Sub, Dataflow, BigQuery, Looker) and ML Platform (Vertex AI, Cloud Storage, BigQuery, Cloud Functions).

• Requirements: application type, expected users and requests per second, budget, team experience, compliance needs and availability targets.
• Architecture: a recommended pattern with service stack, pros and cons, and a rough cost estimate.
• Cost: a breakdown by service, right-sizing, committed and sustained use discount analysis, and potential savings.
• Code: Terraform HCL, gcloud deployment commands, a cloudbuild.yaml, and a security checklist.

How do you install the skill?

Download the ZIP file from this page and unzip it. You can install it in one of two places. For a personal setup available across projects, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills, then upload

What do you need before you start?

The skill's scripts need Python 3. The deployment examples need the Google Cloud SDK (gcloud) and a GCP project where you have permission to run the commands. Terraform is needed only if you use the generated HCL. Docker and either GitHub or Cloud Build are needed for the CI/CD examples.

The source states that the scripts run locally and do not use the network.

• Python 3 (to run the bundled scripts)
• Google Cloud SDK (gcloud) and a GCP project with suitable permissions
• Terraform, if you use the generated HCL
• Docker and a GitHub or Cloud Build setup, for CI/CD

How do you trigger the skill with a natural request?

You do not need to type a script name. Describe what you want in plain language. The skill's description lists the kinds of requests it matches: designing Google Cloud infrastructure, deploying to GKE or Cloud Run, configuring BigQuery pipelines, optimizing GCP costs, or migrating to GCP.

Give the agent as much context as you can in the first message. Application type, expected traffic, budget and team size let it pick a sensible pattern without asking many follow-up questions.

• "Design a Google Cloud setup for a small SaaS web app with a Postgres-compatible database and a budget under a few hundred dollars a month."
• "Review this GCP billing export and suggest right-sizing and discount options."
• "Draft Terraform and a Cloud Build config to deploy my container to Cloud Run."
• "Set up a BigQuery pipeline for clickstream events with Pub/Sub ingestion."

What does a typical session look like?

The agent will ask for requirements, then propose a pattern. The skill's architecture_designer.py script takes a requirements file and returns a recommendation. The cost_optimizer.py script takes a description of your current resources and your monthly spend. The deployment_manager.py script generates deployment scripts for a chosen pattern.

Example commands from the skill:

• python scripts/architecture_designer.py --input requirements.json
• python scripts/cost_optimizer.py --resources current_setup.json --monthly-spend 2000
• python scripts/deployment_manager.py --app-name my-app --pattern serverless_web --region us-central1

What about safety, scripts and network access?

This skill runs scripts. Read the scripts in scripts/ before running them, and run them in a directory where you are comfortable with generated files appearing. The bundled scripts appear to produce illustrative output, so treat their numbers and recommendations as a starting point, not a quote.

The skill's sample outputs use the figures shown in its examples. Check every cost figure against the GCP Pricing Calculator (linked in the skill) and your actual billing data. Review IAM roles, service accounts and secrets against your own security policy before deploying anything.

The security checklist is a general outline. It is not a compliance audit, so do not rely on it alone for HIPAA, SOC 2 or GDPR obligations.

• Verify all generated Terraform and gcloud commands before running them against a real project.
• Test in a dev or staging project first.
• The skill's own failure-handling steps include gcloud logging read and gcloud run services describe; use them to diagnose problems.

Which tips help the most?

Start with the requirements step. The pattern recommendation is only as good as the inputs. The skill's reference files give more detail: architecture_patterns.md for pattern specs, service_selection.md for choosing between services, and best_practices.md for naming, labels, IAM, networking and disaster recovery.

For cost reviews, export your current resource inventory first so the cost_optimizer.py script has real data to analyse. The skill lists idle resource removal, machine type right-sizing, committed use discounts, storage class transitions and network egress optimization as its check areas.

• Read references/service_selection.md before you pick between Cloud Run and GKE.
• Use the naming pattern in best_practices.md, such as prod-myapp-run-api, to keep resources easy to find.
• If a deployment fails, check quota (IAM & Admin > Quotas), IAM permissions and the --allow-unauthenticated flag first.