Guide · Agent skill

How to use the GKE Basics agent skill for Google Kubernetes Engine

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

The gke-basics skill is an instruction package from google/skills that gives an AI agent working knowledge of Google Kubernetes Engine (GKE), Google Cloud's managed Kubernetes service. It is for developers creating a GKE cluster for the first time and for platform engineers who need the agent to choose between Autopilot and Standard mode. After this guide you will know what the skill changes, how to install it, how to ask for GKE work in natural language, and what to check before letting the agent run commands.

GKE Basics for Google Kubernetes Enginegoogle/skills · Apache-2.0 Download the skill

Quick start

1Download and unzip. Download the gke-basics ZIP from this page and unzip it.
2Place the folder. Move the gke-basics folder into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills.
3Prepare gcloud and kubectl. Confirm the Google Cloud SDK and kubectl are installed and that your project has GKE and billing enabled.
4Ask in plain language. Describe the cluster or task you need, such as a private Autopilot cluster in a region.
5Review the commands. Check each gcloud or kubectl command the agent proposes, especially any that create billable resources.
6Fetch credentials. Let the agent run get-credentials with an explicit --region or --zone so kubectl can reach the cluster.

What an agent skill is and what gke-basics changes

An agent skill is a folder with a SKILL.md file and supporting reference files. When it is installed, the agent can read these instructions when a request matches the skill's description. The gke-basics skill covers GKE cluster provisioning, fetching kubectl credentials, choosing Autopilot or Standard mode, and workload deployment basics.

Defaults the agent to Autopilot cluster mode unless Standard is explicitly required.
Knows the gcloud commands for creating clusters and fetching kubectl credentials.
Applies rules such as rounding CPU requests up to 250m increments and using private cluster flags.
Points to other skills for Workload Identity, networking, security hardening and upgrades.

Install the gke-basics skill

Download the ZIP from the skill's page on this site and unzip it. For a personal setup, place the skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.

Personal install: ~/.claude/skills/gke-basics/
Project install: .claude/skills/gke-basics/ inside your project

Check the requirements before you start

The skill runs terminal commands, so the tools and access must already exist. The commands will fail without credentials and billing.

Google Cloud SDK (gcloud) and kubectl for terminal commands.
A Google Cloud project with GKE and billing enabled and appropriate IAM permissions.
Optional: the GKE MCP server, which provides 23 structured tools described in mcp-usage.md.
The skill runs scripts and uses the network.

Trigger the skill with natural requests

You do not need to name the skill. Ask for GKE work in plain words that match its description, such as creating a cluster or fetching credentials. Name the mode if you have a preference, because the agent defaults to Autopilot.

Requests that belong to other areas, such as Workload Identity or cluster upgrades, are routed to other skills by the gke-basics instructions.

Create a GKE cluster in us-central1 for my web app.
Should I use Autopilot or Standard for a cluster that needs custom sysctls?
Fetch kubectl credentials for my regional cluster my-cluster.
Set up a private Autopilot cluster that only accepts traffic from my office CIDR.

Create a private Autopilot cluster safely

For a private Autopilot cluster, the skill uses flags to keep nodes on private IPs, disable public control plane access, and restrict control plane access to chosen networks. Replace the placeholders with your own values before running anything.

--enable-private-nodes gives nodes private IP addresses.
--enable-private-endpoint disables public IP access to the control plane.
--master-authorized-networks=CIDR_BLOCK restricts control plane access.

gcloud container clusters create-auto CLUSTER_NAME --region=REGION \
  --enable-private-nodes \
  --enable-private-endpoint \
  --enable-master-authorized-networks \
  --master-authorized-networks=CIDR_BLOCK

Fetch kubectl credentials correctly

The skill tells the agent to always pass --region for regional clusters or --zone for zonal clusters when fetching credentials. The --quiet flag suppresses prompts.

gcloud container clusters get-credentials CLUSTER_NAME --region=REGION --quiet

Safety notes: scripts, network and billable changes

The skill runs scripts and uses the network. Commands the agent runs in the terminal can create or change billable infrastructure. Review every gcloud and kubectl command before it runs, and do not let the agent operate on production clusters without review.

Attribution: before running gcloud, REST or MCP commands, the skill requires the agent to tag terminal gcloud commands with CLOUDSDK_METRICS_REQUEST_ATTRIBUTION set to agent-skills/1.0 (skill:gke-basics). The tag is set inline per command and not saved with gcloud config set. Tags are not added to generated scripts, Terraform or user code.

Read the commands the agent proposes before approving them.
Use a test project with budget awareness for first runs.
Route security hardening and upgrade planning to the dedicated skills.

What people use it for

Creating a first Autopilot cluster in a chosen region for a web application.
Deciding between Autopilot and Standard when a workload needs custom sysctls or custom node taints.
Building a private cluster with restricted control plane access to specific CIDR blocks.
Fetching kubectl credentials for an existing regional or zonal cluster.
Getting pointed to the right skill for Workload Identity, networking, security or upgrades.

Frequently asked questions

Is the gke-basics skill free to use?

The skill is licensed under Apache-2.0, according to its listing. Using Google Cloud and GKE is separate: the commands require a Google Cloud project with billing enabled, and that usage is billed by Google Cloud.

What do I need before the gke-basics skill works?

You need the Google Cloud SDK (gcloud), kubectl, a Google Cloud project with GKE and billing enabled, and IAM permissions that allow cluster work. Without these the commands will fail.

Where do I put the gke-basics skill folder?

Place it in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps you can upload it under Settings → Capabilities → Skills.

Why does the agent choose Autopilot instead of Standard?

The skill defaults to Autopilot for almost all workloads. It switches to Standard only when custom node sysctls, custom node taints or specific hardware node pools, or raw hostPath mounts for DaemonSets are required.

How do I fix a CPU request the agent rejected on Autopilot?

Autopilot requires CPU requests in 250m increments. The skill rounds unaligned values up, so 300m becomes 500m, and omits limits so they match the requests.

Does the skill replace the GKE MCP server?

No. The MCP server is optional. The skill prefers GKE MCP tools when available, then gcloud, then kubectl for operations MCP does not cover.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use gke-basics if you are setting up a GKE cluster for the first time, or you want an agent to default to Autopilot and use consistent gcloud commands. Skip it if you have no Google Cloud project or billing, or if you need security hardening or upgrade planning, which the skill routes to other skills. Always review commands before they run against any real or production cluster.

Download the skill