GKE Workload Security
Audit and harden workload-level security controls for GKE applications and namespaces.
The skill covers Workload Identity, network isolation, read-only secret mounts, and least-privilege guidance, with no malware or covert exfiltration indicators; however, several commands directly mutate a cluster without explicit per-action confirmation or rollback, and the operational impact of default-deny policies and namespace labeling is not fully disclosed, so points are deducted.
The audit script has argument checks, set -e, and pipefail, and prerequisites name gcloud and jq; however, there are no committed tests or abnormal-case demonstrations. The workload identity example sets automountServiceAccountToken to false, which may impair reproduction, and failure diagnostics and environment prerequisites are limited. Static calibration therefore keeps this at 8.
Invocation scenarios, covered areas, and non-fit areas are relatively clear, spanning audits, identity, network policy, sandboxing, Pod Security, and CSI; however, input/output contracts, version compatibility boundaries, Chinese-language support, and mainland-China network considerations are absent, and some controls depend on specific GKE configurations, warranting deductions.
The documentation is organized by workflow and includes scripts, YAML assets, prerequisites, examples, official resources, and repository-level Apache-2.0 licensing and installation guidance; however, the skill lacks its own versioning, changelog, explicit maintenance owner, troubleshooting section, and update path, while some dependencies remain underspecified.
The audit script and several directly usable commands/manifests cover core security configuration tasks and provide practical value; however, complete verification, policy exception handling, rollback, and production adaptation guidance are missing. Default-deny enforcement may require substantial additional configuration, so the static score remains conservative.
The skill provides auditable source code, a pinned image digest, configuration assets, and official documentation references, enabling limited source review; however, it contains no CI, test suite, execution results, or third-party reproduction evidence, so key paths are not independently verifiable.
- Before running gcloud, kubectl, or namespace-label commands, confirm the target cluster, region, project, change window, and recovery procedure.
- A default-deny network policy can interrupt existing ingress, egress, DNS, or cross-namespace traffic; inventory dependencies and roll out incrementally.
- Validate automountServiceAccountToken=false in the Workload Identity example, the Secret Manager CSI prerequisite, and GKE version compatibility in the target environment.
- The documentation does not address mainland-China reachability, proxies, or enterprise network requirements; confirm access paths before relying on Google Cloud services and documentation.
What does this skill do, and when should you use it?
This skill provides operational workflows for securing Google Kubernetes Engine workloads across identity, networking, pod isolation, pod security, and secret mounting. It can run a cluster audit script and configure Workload Identity Federation, Network Policies, GKE Sandbox, Pod Security Standards, and Secret Manager CSI volumes. It explicitly excludes control-plane security, RBAC hardening, Binary Authorization, Shielded Nodes, and platform-level GKE add-ons. It fits teams that can execute authenticated gcloud, kubectl, and local shell commands against a GKE environment.
Runs scripts/audit_cluster.sh to check Workload Identity, Network Policy, Shielded Nodes, Binary Authorization, and private-cluster configuration; creates namespaces and Kubernetes Service Accounts and binds and annotates Google Service Accounts; enables or applies Network Policies and Dataplane V2 NetworkLogging; sets runtimeClassName: gvisor for sandboxed Pods; enforces the restricted Pod Security Standard through namespace labels; and supplies SecretProviderClass and Pod CSI volume examples for mounting Secret Manager secrets.
- A GKE platform engineer needs to audit a cluster's workload security posture with gcloud and jq.
- An application team needs a Kubernetes Service Account to access Google Cloud APIs through Workload Identity.
- A multi-tenant cluster administrator needs to apply default-deny ingress and egress policies to a namespace.
- A security engineer needs to isolate untrusted workloads with GKE Sandbox and enforce the restricted Pod Security Standard.
- A DevOps team needs to mount Google Secret Manager values as read-only volumes through the Secret Manager CSI Driver.
What are this skill's strengths and limitations?
- Covers practical workload controls for identity, network isolation, pod isolation, pod security, and secret volumes.
- Includes copyable gcloud, kubectl, YAML, and audit-script examples.
- Clearly separates workload-level responsibilities from the platform-level scope assigned to gke-platform-security.
- Includes Dataplane V2 network logging guidance and official Google Cloud resource links.
- Requires an authenticated gcloud CLI, jq, and kubectl; the source does not specify the complete IAM permission set.
- The source provides no test suite, compatibility matrix, or sample audit output.
- The Network Policy enablement command is unnecessary and may fail on Dataplane V2 clusters.
- It does not cover control-plane security, RBAC hardening, Binary Authorization, Shielded Nodes, or platform-level add-on enablement.
How do you install this skill?
Install it as part of the google/skills collection with: npx skills add google/skills. The installer allows selection of individual skills from the repository. The target file is skills/cloud/gke-workload-security/SKILL.md; the README does not document the exact installation directory or a non-interactive selection parameter.
How do you use this skill?
In an Agent Skills-compatible client, request a workload-security task such as: “Audit this GKE cluster and configure Workload Identity and a default-deny Network Policy for the target namespace.” Before running the audit, authenticate the gcloud CLI and install jq; the configuration workflows also use kubectl. The documented Network Policy command is: gcloud container clusters update <cluster-name> --update-addons=NetworkPolicy=ENABLED --region <region>.
How does this skill compare with similar options?
Compared with gke-platform-security, this skill focuses on application, Pod, and namespace controls. The source places platform-level actions such as enabling Shielded Nodes and GKE Sandbox in that separate skill.