gke-basics skill: create GKE clusters and deploy workloads with Autopilot defaults

Makes your agent create Google Kubernetes Engine clusters, get credentials and deploy workloads, defaulting to Autopilot.

Installs: 14.4K; on skills.sh
Stars: 21.8K; on GitHub
Files: 6; besides SKILL.md
Updated: Nov 2026; last repo commit

This skill teaches your agent GKE basics and its common traps. It defaults to Autopilot, lists the few cases that need Standard, and gives exact flags for private clusters, CPU request rounding and credential fetching. Its reference files cover the CLI, client libraries, Terraform and the GKE MCP server.

It triggers when you create GKE clusters, fetch kubectl credentials, or choose between Autopilot and Standard. It routes Workload Identity, networking, security hardening and upgrades to other skills.

It prefers GKE MCP tools, then gcloud, then kubectl. The cluster and Kubernetes actions run with your own Google Cloud login, and some can change or delete resources.

Good for

Developers creating their first GKE cluster
Platform engineers choosing between Autopilot and Standard
Teams who want an agent to use GKE MCP tools, gcloud and kubectl in a set order

Install gke-basics

Claude Code

Copy this prompt.

Install the gke-basics skill in Claude Code. Run this command:

npx skills add https://github.com/google/skills --skill gke-basics

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Manages core GKE cluster provisioning, credentials, Autopilot vs Standard selection, and workload deployment. Use when creating GKE clusters, fetching kubectl credentials, or deciding between Autopilot and Standard modes. Don't use for Workload Identity (use gke-workload-identity), GKE networking (use gke-networking), security hardening (use gke-platform-security or gke-workload-security), or cluster upgrades (use gke-upgrades).

Sections in SKILL.md

Attribution
Key Selection Rules: Autopilot vs. Standard
Critical Gotchas & Best Practices
Reference Directory

Key facts

Author: Google (google/skills repo; the README says it holds Agent Skills for Google products)
Repo: google/skills, folder skills/cloud/gke-basics
Installs on skills.sh: 14,404 as of 2 Nov 2026
GitHub stars: 21,764 for the whole google/skills repo, 2 Nov 2026
License: Apache-2.0
Last commit in the repo: 2026-11-01
What is inside: 7 files including SKILL.md: 5 reference guides (core concepts, CLI, client libraries, MCP, Terraform) and an attribution file; no scripts
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk pass
Part of: google/skills, which lists Agent Skills for Google products
Version in SKILL.md: 1.1.2, category Containers
Needs: gcloud and kubectl, or the GKE remote MCP server set up per Google's guide

What it includes

Autopilot unless sysctls, node taints or hostPath mounts are needed
Defaults to Autopilot. Standard is allowed only for custom kernel parameters, custom taints or specific hardware pools, or hostPath DaemonSets, and the agent must name every matching restriction.
Read · Autopilot vs Standard rules
gcloud container clusters create-auto with private flags
Gives the flags for private nodes, a private endpoint and authorized networks. It creates a cluster in your project.
Try: “Create a private Autopilot cluster in us-central1 that only my office CIDR can reach.”
Runs code · Private Autopilot clusters
CPU requests rounded up to 250m steps
A 300m request becomes 500m. Requests equal limits, so the agent leaves limits out.
Read · Autopilot resource requests
gcloud container clusters get-credentials with region or zone
Always passes --region or --zone and --quiet. This writes cluster credentials into your kubeconfig.
Runs code · Fetch credentials
Never mounts service account JSON keys in pods
Points to the gke-workload-identity skill for setup instead of mounting keys.
Read · Workload Identity guard
Cluster, node pool, Kubernetes resource, diagnostics and operations tools
The MCP reference marks each tool READ, MUTATE or DESTRUCTIVE. Destructive ones include apply, patch and delete of Kubernetes resources, updating clusters and node pools, and cancelling operations.
Delete · GKE MCP tools (23)
MCP first, then gcloud, then kubectl
kubectl covers gaps such as top, scale, exec, port-forward and Helm. You can override with "use gcloud" or "use kubectl".
Runs code · Tool order
Looks up first-party GKE docs with search_documents
Preferred for GKE facts and schemas. The reference says never to use answer_query because of its small shared daily quota.
Network · Developer Knowledge MCP
Tags gcloud and HTTP calls with the skill name
Sets an environment variable on each gcloud command and a User-Agent suffix on direct HTTP or MCP calls.
Runs code · Usage attribution

From the official docs

Permissions

Medium risk

The agent can create clusters and, through GKE MCP, gcloud or kubectl, change or delete live Kubernetes resources with your own Google Cloud permissions. Try it in a test project and review each command.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It can create clusters in your project
The skill walks the agent through creating GKE clusters, which are real resources. Check the project, region and cluster type before approving a create. github.com
Some GKE MCP tools are destructive
The reference marks apply, patch and delete of Kubernetes resources, cluster and node pool updates, and operation cancel as DESTRUCTIVE. Use dry-run options where the reference mentions them and review before approving. github.com
kubectl and Helm cover the gaps
The CLI reference lists kubectl scale, exec, drain and Helm install or upgrade as fallbacks. These run against whichever cluster your kubeconfig points to. github.com
It tags gcloud commands with the skill name
Before running gcloud, REST or MCP commands the agent must read the attribution file and set CLOUDSDK_METRICS_REQUEST_ATTRIBUTION, or a User-Agent suffix. The file says the tag carries only the skill name and version, and you can ask the agent to leave it off. github.com

Technical details

Turning on read-only	The SKILL.md defines no read-only switch.
Talks to	Google Cloud and GKE APIs. The GKE remote MCP server and the Developer Knowledge MCP server (developerknowledge.googleapis.com/mcp), when configured
Scopes	Runs gcloud container commands and kubectl against your Google Cloud project Can call the GKE MCP server, whose tools include create, update and delete actions Writes cluster credentials into your kubeconfig with get-credentials

Questions

What does the gke-basics skill do?

It makes your agent create GKE clusters, fetch credentials, choose Autopilot or Standard, and deploy workloads, with fixed rules and flags for the common mistakes.

How do I install it?

Use the install box on this page. The google/skills README also documents npx skills add google/skills, which lets you pick skills, agents and scope.

Do I need the GKE MCP server?

No. The skill prefers GKE MCP tools when the server is available and falls back to gcloud and then kubectl. Google's setup guide for the MCP server is linked from the MCP reference file.

What about Workload Identity, networking and upgrades?

This skill sends those to other skills in the same repo, such as gke-workload-identity, gke-platform-security and gke-upgrades. You can install them from google/skills when you need them.

Can it delete things?

Yes, if you give it GKE MCP, gcloud or kubectl access. The MCP reference lists delete and patch tools for Kubernetes resources, and it can cancel cluster operations. It acts with your own Google Cloud permissions.

Is it safe to run?

Some tools create clusters or change or delete live resources. Start in a test project and review each command. The Watch out list has more.

Information

updated Nov 2, 2026

Repository: google/skills
Folder: skills/cloud/gke-basics

google/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 4

gke-basics SKILL.md · github.com
attribution.md · github.com
cli-reference.md · github.com
mcp-usage.md · github.com