GKE App Onboarding Assistant
Containerize and deploy an application to Google Kubernetes Engine for the first time.
The skill includes non-root execution, a read-only root filesystem, disabled privilege escalation, resource limits, vulnerability scanning, and digest-pinned image guidance. However, it does not require confirmation before applying Kubernetes changes, explain credentials, secrets, or data flows, or provide rollback and recovery procedures, so points are deducted.
The workflow is internally organized across assessment, build, image storage, manifest generation, deployment, and verification. However, the main example uses /readyz while the supplied Node example implements only /healthz; MCP parameters and abnormal-failure feedback are underspecified, and no visible tests cover key paths, so points are deducted.
The audience and triggering scenario are clear, and general cluster administration and upgrades are explicitly out of scope. However, input and output contracts, migration boundaries, non-trigger conditions, Chinese-language support, mainland-China reachability, and broader stack coverage are not documented, so points are deducted.
The material uses a sensible progression from assessment through deployment, includes assets, and benefits from repository-level Apache-2.0 licensing, installation guidance, and support links. It lacks skill-specific versioning, changelog, maintenance ownership, prerequisite documentation, and troubleshooting guidance, so points are deducted.
The skill can guide first-time containerization and basic GKE deployment, with a reusable Node example and Kubernetes manifest. However, migration details, secrets, external dependencies, production exposure, and rollback still require substantial manual judgment; static evidence cannot establish dependable end results, so points are deducted.
The supplied source, example code, and manifest are auditable, and the repository revision is specified. However, there is no committed key-path test suite, CI execution evidence, or third-party reproduction evidence for this skill, so verifiability remains limited.
- Before running apply_k8s_manifest or kubectl apply, show the proposed changes, obtain user confirmation, and prepare an actionable rollback plan.
- Resolve the health-check contract mismatch: the main document requires /readyz while the supplied Node example provides only /healthz; also document secrets, identity permissions, network exposure, and sensitive-data handling.
- Do not use the example LoadBalancer or default namespace unchanged in production; reassess exposure, cost, and organizational policy.
What does this skill do, and when should you use it?
This skill supports first-time application onboarding to GKE, covering assessment, containerization, image storage, Kubernetes manifest generation, deployment, and verification. It examines the stack, dependencies, configuration, state, networking, and health endpoints before deployment. GKE Kubernetes MCP tools are preferred, with kubectl commands available as a fallback. It is not intended for general GKE cluster administration or upgrades.
It assesses an application's language and framework, dependencies, configuration, persistence needs, network protocol, and health endpoints; provides Dockerfile and Cloud Native Buildpacks examples; configures Artifact Registry authentication with gcloud and builds and pushes images with docker; generates Kubernetes YAML containing a Deployment, Service, resource requests and limits, liveness and readiness probes; deploys through apply_k8s_manifest or kubectl; and checks rollout status and Pods.
- A developer is containerizing an existing application for its first GKE deployment.
- A platform engineer needs baseline Kubernetes Deployment and Service manifests for an application.
- A team has pushed an image to Artifact Registry and needs to verify its GKE rollout and Pods.
- A migration project needs to inventory configuration, storage, ports, and health checks before moving an application to GKE.
What are this skill's strengths and limitations?
- Covers the core path from application assessment through first deployment verification.
- Includes concrete practices such as multi-stage builds, non-root execution, and minimal base images.
- Provides manifest checks for resource sizing, health probes, and replica count.
- Offers both MCP-tool and kubectl deployment paths.
- Its scope is first-time application onboarding and deployment, not general cluster administration or upgrades.
- It provides no test suite, automation scripts, or complete troubleshooting procedure.
- Specific IAM permissions, costs, and supported platform versions are not documented.
- Vulnerability scanning examples cover enabling and querying scans, not remediation.
How do you install this skill?
Run npx skills add google/skills and select the gke-app-onboarding skill during installation. The source does not document the exact installation directory.
How do you use this skill?
Ask a compatible Agent Skills client to do something like “Onboard this application to GKE for the first time,” and provide the application code plus project, region, cluster, Artifact Registry repository, and image details. Prefer apply_k8s_manifest, then verify with get_k8s_rollout_status and get_k8s_resource; alternatively use kubectl apply -f manifests/, kubectl rollout status deployment/my-app, and kubectl get pods -l app=my-app.
How does this skill compare with similar options?
Its scope differs from gke-basics and gke-upgrades: this skill focuses on first-time application containerization, manifest generation, and deployment, while the source directs general cluster administration to gke-basics and upgrades to gke-upgrades. For deployment execution, MCP tools are preferred and kubectl is the fallback.