How to use the Global Load Balancer Setup agent skill

What the skill makes the agent do

A global external Application Load Balancer sits in front of your websites or APIs and spreads traffic across backends. This skill guides the agent through a 6-step discovery process to design that setup, mapping your workload to opinionated best-practice configurations that use actual Google Cloud product names.

The agent asks questions one at a time: project and basics, origins (Cloud Storage buckets, Managed Instance Groups, GKE clusters, Cloud Run services, or external IP/FQDN origins), routing rules, traffic management, caching with Cloud CDN, and security with Cloud Armor. It then generates Terraform HCL or gcloud CLI scripts, and can help actuate deployments or check for configuration drift.

• Caching is proposed from your workload type: static objects, cacheable API, uncacheable API or transactions, or dynamic web (SSR).
• Security settings such as rate limiting and OWASP protection are also proposed from the workload type, with advanced options hidden unless you ask.
• Drift detection uses Infrastructure Manager previews and presents a choice: reconcile to code, or backport manual changes into the Terraform code.

Install the skill from the ZIP

Download the ZIP file from this page and unzip it. To make the skill available to every project on your machine, unzip it into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder.

You can also upload the skill in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the ZIP there. The skill package includes SKILL.md and a references/ folder with attribution.md, drift-detection.md, gcloud-generation.md, managed-deployment.md, resource-discovery.md, terraform-generation.md and terraform-module.md.

Trigger the skill with a natural request

You do not need a special command. The skill's description tells the agent when to use it, so describe what you are trying to do. Requests such as designing or deploying a global external Application Load Balancer, adding Cloud CDN or Cloud Armor, generating Terraform for a load balancer, or checking a deployed load balancer for drift should match.

Be specific about your goal and, if you have them, name the backends you already run. The agent will then discover existing resources where it can.

• Example: Set up a global external load balancer in front of my Cloud Run service with Cloud CDN and Cloud Armor.
• Example: Generate Terraform for a global load balancer that serves static images from a Cloud Storage bucket.
• Example: Check my existing global load balancer for configuration drift and show me the differences.
• Example: Help me add Cloud Armor rate limiting to my API load balancer.

What you need before you start

The skill is designed to run where the Google Cloud CLI (gcloud) can be executed. gcloud is needed for automated resource discovery and managed deployment. Without it, the skill only supports design and generating Terraform HCL.

You also need a Google Cloud project with permissions to read and change load balancing, IAM and related resources. For deployment you need Terraform or Infrastructure Manager. The skill says it can run gcloud commands and deploy resources, which affect real traffic and cost money.

• Google Cloud CLI (gcloud) for discovery and deployment
• A Google Cloud project with the required permissions
• Terraform or Infrastructure Manager if you deploy through those tools

Safety notes: scripts and network access

This skill runs scripts and uses the network. Before any gcloud, REST API or MCP command, the agent is required to read the attribution reference. Terminal gcloud commands are prefixed with a CLOUDSDK_METRICS_REQUEST_ATTRIBUTION variable that identifies the skill and does not carry user data, and the source says gcloud config set should not be used for this.

Drift detection is designed to be safe: the skill's guidance says never to modify production infrastructure directly during drift detection and to generate a preview first. Still, review generated scripts and Terraform before running them against production, because deployments change live traffic.

• Generated scripts and Terraform should be reviewed before they touch production.
• The skill excludes non-Google Cloud load balancers and purely regional or internal load balancers unless they are part of a global failover design.
• The full content of the reference files was not included in the source, so their exact behaviour is only partly verified.

Tips for getting good results

Answer each question in the order the agent asks it. The flow is deliberately sequential, and the agent waits for your response before moving on. If you want control over caching or security details, say so explicitly, because advanced settings stay hidden by default.

Start with a design or Terraform generation request if you lack gcloud access or are not ready to deploy.

• Name your origin types clearly so the agent can discover matching resources.
• Choose the workload type carefully, since it drives both caching and security defaults.
• Ask for a drift check on an existing load balancer before changing it by hand.