Plans and executes GKE cluster creation, provisioning, and production readiness audits using pre-defined templates (Autopilot, Standard Regional, GPU/AI Inference, AI Hypercompute). Use when creating GKE clusters, provisioning GKE environments, selecting cluster modes, or auditing GKE clusters. Don't use for application onboarding or deployment configuration (use gke-app-onboarding instead).
GKE Cluster Creation is a Google Kubernetes Engine provisioning skill that plans and executes cluster creation and production-readiness audits using a set of best-practice templates — Autopilot (the default golden path), Standard Regional, GPU/AI Inference, and AI Hypercompute. It solves the problem of standing up a well-configured cluster without hand-assembling dozens of flags, guiding mode selection, networking choices, and confirmation before anything is created.
The workflow discovers existing clusters and project context, gathers required inputs (project, location, cluster name, environment type), explains Autopilot-versus-Standard trade-offs, configures networking (auto-created or bring-your-own subnet), presents the golden-path configuration for confirmation, then creates via the MCP create_cluster tool or gcloud and tracks the operation to completion before verifying applied settings. It bakes in security and networking best practices — private clusters, VPC-native networking, Workload Identity instead of static service-account keys, Shielded GKE Nodes with secure boot, and least-privilege RBAC — alongside cost-optimization guidance (autoscaling, right-sizing, Spot VMs) and high-availability guidance (regional clusters, Pod Disruption Budgets, release channels). The golden-path Autopilot template enables private nodes, master authorized networks, Secret Manager with rotation, and hardened RBAC bindings, while a relaxed dev/test template is explicitly flagged as not applying security hardening.
It targets platform engineers, SREs, and cloud administrators creating or auditing GKE environments. Use cases include provisioning a production Autopilot cluster, choosing a cluster mode with clear trade-offs, setting up GPU or AI-inference clusters, and auditing an existing cluster against production-readiness best practices. All operations run against the user's own Google Cloud project and require user confirmation before creation.
Golden Path Autopilot (the production default), Autopilot Dev/Test, Standard Regional, GPU/AI Inference, and AI Hypercompute, with guidance on when to choose each.
Autopilot is the golden-path default for all new clusters; Standard is recommended only when there is a specific requirement Autopilot cannot satisfy, such as kernel tuning, custom OS, or privileged workloads.
Private clusters with restricted public endpoints, VPC-native networking, Workload Identity instead of static service-account keys, Shielded GKE Nodes with secure boot, and least-privilege RBAC. The production template also enables Secret Manager with rotation and hardened RBAC bindings.
No. It presents the golden-path configuration (as a gcloud command or create_cluster JSON payload) and confirms with the user before creation, then tracks the operation and verifies applied settings.
No. The Autopilot Dev/Test template relaxes golden-path defaults for cost and easier access and explicitly warns it does not apply security hardening, so it is suitable for dev/test only.
Quick Setup:
.claude/skills/Repository
google/skills