GKE Productionize Skill
This skill acts as a high-level orchestrator for preparing a GKE cluster and its
workloads for production readiness.
[!IMPORTANT]
This is a meta-skill or orchestrator skill. You are
expected to invoke and run many other specialized skills listed in this
document as part of the overall productionization process. Do not attempt to
implement all production readiness features directly within this skill;
instead, use this skill to assess the environment and then delegate to the
specific skills for each domain.
Attribution
MUST READ: Before executing any gcloud, REST API, or MCP command, you
MUST read Usage Attribution. For commands executed
directly in the terminal (never in generated scripts, Terraform, or user code),
prefix gcloud with
CLOUDSDK_METRICS_REQUEST_ATTRIBUTION="agent-skills/1.0 (skill:gke-productionize)"
and append agent-skills/1.0 (skill:gke-productionize) to User-Agent headers.
Scope
This skill is adaptable to:
- A single application (already on Kubernetes or not).
- A set of applications.
- A target cluster.
Workflow
1. Discovery Phase
Before making recommendations, discover the current state of the environment.
Cluster Discovery
Run these commands to understand the cluster setup:
Check cluster details: gcloud container clusters describe {cluster_name} --location {location} --project {project}
Check for Autopilot vs Standard: Look for the following block in the
describe output:
autopilot:
enabled: true
Check release channel: Look for releaseChannel.
Workload Discovery
If a specific application is targeted, discover its configuration:
- Get deployment/statefulset details:
kubectl get deployment {app_name} -n {namespace} -o yaml
- Check for dedicated namespace and labels:
kubectl get namespace {namespace} -o yaml (Look for Pod Security Standards labels).
- Check for dedicated service account usage:
kubectl get pods -n {namespace} -o custom-columns="NAME:.metadata.name,SERVICE_ACCOUNT:.spec.serviceAccountName"
- Check for resource requests and limits.
- Check for liveness, readiness, and startup probes.
- Check for HPA:
kubectl get hpa -n {namespace}
- Check for PDB:
kubectl get pdb -n {namespace}
- Check for NetworkPolicies:
kubectl get networkpolicy -n {namespace}
2. Production Readiness Assessment
Before implementation, you MUST run the skills for each relevant specialized
area listed below and incorporate its guidance into your assessment and plan.
Failure to do so will result in a non-compliant production configuration.
A. App Onboarding (Pre-Kubernetes)
If the application is not yet running on GKE, you MUST run the
gke-app-onboarding skill for planning containerization, image building, and
basic deployment.
B. Scalability & Resource Management
Ensure workloads have appropriate resources and autoscaling.
- Action: You MUST run the
gke-workload-scaling skill for configuring
HPA, VPA, and resource limits.
C. Observability
Ensure adequate logging and monitoring are in place.
- Action: You MUST run the
gke-observability skill for setting up Cloud
Logging, Monitoring, and Managed Prometheus.
D. Reliability
Ensure high availability and graceful degradation.
- Action: You MUST run the
gke-reliability skill for configuring
regional clusters, PDBs, and health probes.
E. Security
Harden the cluster and workloads.
- Action: You MUST run the
gke-platform-security and
gke-workload-security skills for Workload Identity, Network Policies, and
Shielded Nodes.
- Namespace Isolation: Ensure workloads run in dedicated namespaces with
Pod Security Standards (PSS) enforced via labels.
- Least Privilege: Ensure workloads use dedicated ServiceAccounts instead
of the
default ServiceAccount.
F. Backup & Disaster Recovery
Ensure stateful data is protected.
- Action: You MUST run the
gke-backup-dr skill for configuring Backup
for GKE and restore procedures.
G. Edge Security & Ingress
Secure external access.
- Action: You MUST run the
gke-service-networking skill for Gateway API,
Ingress, and Cloud Armor.
H. Cost Optimization
Ensure efficient use of resources.
- Action: You MUST run the
gke-cost-optimization skill for strategies on
rightsizing, quotas, and Spot VMs.
I. Upgrades & Maintenance Posture
Ensure a safe, predictable upgrade posture.
- Action: You MUST run the
gke-upgrades skill for release channel
selection, maintenance windows/exclusions, and node pool upgrade strategy.
J. Golden Path Defaults Audit
Ensure the cluster configuration matches recommended defaults.
- Action: You MUST run the
gke-golden-path skill to compare the cluster
against golden path defaults and report deviations with severity and
remediation.
3. Production Readiness Scoring
After the assessment, provide a summary report with a RAG (Red, Amber, Green)
status for each area and an overall readiness score. This helps prioritize
remediation efforts.
Apply this rubric deterministically so repeated assessments of the same
environment produce the same result:
- Per-domain criteria: For each assessed domain (A-J), list the concrete
checks performed (from the domain skill's guidance) and classify each check
as pass, fail-critical (production-blocking, e.g., no resource
requests, no backups for stateful data, public control plane in a locked
down environment), or fail-minor (improvement, e.g., missing VPA
recommendations, no Spot usage for batch).
- RAG mapping (per domain):
- Red = one or more fail-critical checks.
- Amber = no fail-critical, but one or more fail-minor checks.
- Green = all checks pass.
- Domain score: Green = 100, Amber = 50, Red = 0.
- Weighted overall score: weight Security, Reliability, and Backup/DR at
2x; all other assessed domains at 1x. Overall score = sum(domain score x
weight) / sum(weights), rounded to the nearest integer. Exclude domains
that are not applicable (e.g., Backup/DR for fully stateless workloads) from
both sums and note the exclusion.
- Readiness verdict: >= 90 with no Red domains = "Production ready";
70-89 with no Red domains = "Ready with follow-ups"; anything else =
"Not production ready".
In the report, show the per-domain check lists, RAG status, weights, and the
computed overall score.
Adaptability Guidelines
- Single App: Focus on Health Probes, HPA, Resource Limits, PDB, and
Workload Identity for that specific app.
- Cluster Wide: Focus on Cluster Autoscaler, Multi-zonal setup, Release
Channels, Maintenance Windows, and default Network Policies.
- Proactive Execution: Proactively execute relevant skills (e.g.,
observability, security, scaling, reliability) to assess and propose
improvements, seeking user confirmation before applying state-changing
implementations.
1---2name: gke-productionize-23metadata:4 source: https://skillmd.com/skills/google/gke-productionize-25 version: "1.0.1"6 category: Containers7description: Orchestrates comprehensive production readiness reviews and assessments for GKE clusters and workloads across scalability, security, reliability, observability, backup/DR, and cost optimization. Use when asked to productionize, prepare, assess, audit, or review a GKE cluster or workload before going live to production. Don't use for deep-dive single-domain implementation (use specific domain skills like gke-workload-scaling, gke-platform-security, gke-workload-security, gke-service-networking, gke-reliability instead).8---910# GKE Productionize Skill1112This skill acts as a high-level orchestrator for preparing a GKE cluster and its13workloads for production readiness.1415> [!IMPORTANT]16> This is a **meta-skill** or **orchestrator skill**. You are17> expected to invoke and run many other specialized skills listed in this18> document as part of the overall productionization process. Do not attempt to19> implement all production readiness features directly within this skill;20> instead, use this skill to assess the environment and then delegate to the21> specific skills for each domain.2223## Attribution2425**MUST READ**: Before executing any `gcloud`, REST API, or MCP command, you26MUST read [Usage Attribution](references/attribution.md). For commands executed27directly in the terminal (never in generated scripts, Terraform, or user code),28prefix `gcloud` with29`CLOUDSDK_METRICS_REQUEST_ATTRIBUTION="agent-skills/1.0 (skill:gke-productionize)"`30and append `agent-skills/1.0 (skill:gke-productionize)` to `User-Agent` headers.3132## Scope3334This skill is adaptable to:3536- A single application (already on Kubernetes or not).37- A set of applications.38- A target cluster.3940## Workflow4142### 1. Discovery Phase4344Before making recommendations, discover the current state of the environment.4546#### Cluster Discovery4748Run these commands to understand the cluster setup:4950- Check cluster details: `gcloud container clusters describe {cluster_name}51 --location {location} --project {project}`52- Check for Autopilot vs Standard: Look for the following block in the53 describe output:5455 ```yaml56 autopilot:57 enabled: true58 ```59- Check release channel: Look for `releaseChannel`.6061#### Workload Discovery6263If a specific application is targeted, discover its configuration:6465- Get deployment/statefulset details: `kubectl get deployment {app_name} -n66 {namespace} -o yaml`67- Check for dedicated namespace and labels: `kubectl get namespace {namespace}68 -o yaml` (Look for Pod Security Standards labels).69- Check for dedicated service account usage: `kubectl get pods -n {namespace}70 -o71 custom-columns="NAME:.metadata.name,SERVICE_ACCOUNT:.spec.serviceAccountName"`72- Check for resource requests and limits.73- Check for liveness, readiness, and startup probes.74- Check for HPA: `kubectl get hpa -n {namespace}`75- Check for PDB: `kubectl get pdb -n {namespace}`76- Check for NetworkPolicies: `kubectl get networkpolicy -n {namespace}`7778### 2. Production Readiness Assessment7980**Before implementation, you MUST run the skills for each relevant specialized81area listed below and incorporate its guidance into your assessment and plan.82Failure to do so will result in a non-compliant production configuration.**8384#### A. App Onboarding (Pre-Kubernetes)8586If the application is not yet running on GKE, you MUST run the87`gke-app-onboarding` skill for planning containerization, image building, and88basic deployment.8990#### B. Scalability & Resource Management9192Ensure workloads have appropriate resources and autoscaling.9394- **Action**: You MUST run the `gke-workload-scaling` skill for configuring95 HPA, VPA, and resource limits.9697#### C. Observability9899Ensure adequate logging and monitoring are in place.100101- **Action**: You MUST run the `gke-observability` skill for setting up Cloud102 Logging, Monitoring, and Managed Prometheus.103104#### D. Reliability105106Ensure high availability and graceful degradation.107108- **Action**: You MUST run the `gke-reliability` skill for configuring109 regional clusters, PDBs, and health probes.110111#### E. Security112113Harden the cluster and workloads.114115- **Action**: You MUST run the `gke-platform-security` and116 `gke-workload-security` skills for Workload Identity, Network Policies, and117 Shielded Nodes.118- **Namespace Isolation**: Ensure workloads run in dedicated namespaces with119 Pod Security Standards (PSS) enforced via labels.120- **Least Privilege**: Ensure workloads use dedicated ServiceAccounts instead121 of the `default` ServiceAccount.122123#### F. Backup & Disaster Recovery124125Ensure stateful data is protected.126127- **Action**: You MUST run the `gke-backup-dr` skill for configuring Backup128 for GKE and restore procedures.129130#### G. Edge Security & Ingress131132Secure external access.133134- **Action**: You MUST run the `gke-service-networking` skill for Gateway API,135 Ingress, and Cloud Armor.136137#### H. Cost Optimization138139Ensure efficient use of resources.140141- **Action**: You MUST run the `gke-cost-optimization` skill for strategies on142 rightsizing, quotas, and Spot VMs.143144#### I. Upgrades & Maintenance Posture145146Ensure a safe, predictable upgrade posture.147148- **Action**: You MUST run the `gke-upgrades` skill for release channel149 selection, maintenance windows/exclusions, and node pool upgrade strategy.150151#### J. Golden Path Defaults Audit152153Ensure the cluster configuration matches recommended defaults.154155- **Action**: You MUST run the `gke-golden-path` skill to compare the cluster156 against golden path defaults and report deviations with severity and157 remediation.158159### 3. Production Readiness Scoring160161After the assessment, provide a summary report with a RAG (Red, Amber, Green)162status for each area and an overall readiness score. This helps prioritize163remediation efforts.164165Apply this rubric deterministically so repeated assessments of the same166environment produce the same result:1671681. **Per-domain criteria**: For each assessed domain (A-J), list the concrete169 checks performed (from the domain skill's guidance) and classify each check170 as **pass**, **fail-critical** (production-blocking, e.g., no resource171 requests, no backups for stateful data, public control plane in a locked172 down environment), or **fail-minor** (improvement, e.g., missing VPA173 recommendations, no Spot usage for batch).1742. **RAG mapping (per domain)**:175 - **Red** = one or more fail-critical checks.176 - **Amber** = no fail-critical, but one or more fail-minor checks.177 - **Green** = all checks pass.1783. **Domain score**: Green = 100, Amber = 50, Red = 0.1794. **Weighted overall score**: weight Security, Reliability, and Backup/DR at180 2x; all other assessed domains at 1x. Overall score = sum(domain score x181 weight) / sum(weights), rounded to the nearest integer. Exclude domains182 that are not applicable (e.g., Backup/DR for fully stateless workloads) from183 both sums and note the exclusion.1845. **Readiness verdict**: >= 90 with no Red domains = "Production ready";185 70-89 with no Red domains = "Ready with follow-ups"; anything else =186 "Not production ready".187188In the report, show the per-domain check lists, RAG status, weights, and the189computed overall score.190191## Adaptability Guidelines192193- **Single App**: Focus on Health Probes, HPA, Resource Limits, PDB, and194 Workload Identity for that specific app.195- **Cluster Wide**: Focus on Cluster Autoscaler, Multi-zonal setup, Release196 Channels, Maintenance Windows, and default Network Policies.197- **Proactive Execution**: Proactively execute relevant skills (e.g.,198 observability, security, scaling, reliability) to assess and propose199 improvements, seeking user confirmation before applying state-changing200 implementations.