What does this skill do, and when should you use it?
This skill targets Google Kubernetes Engine and covers Cloud Logging, Cloud Monitoring, Managed Prometheus, and Dataplane V2 flow metrics. It defines a golden-path observability configuration that includes control-plane metrics and documents workflows using gcloud, kubectl, Logging Query Language, and MCP tools. It also covers alerting, node-health assessment, audit logs, cost control, and production recommendations for Cloud Trace and Cloud Profiler. It is not intended for local application logging frameworks or external APMs outside GKE.
Uses GKE MCP tools to retrieve cluster data, Kubernetes events, logs, cluster information, and resource descriptions; uses gcloud container clusters update to enable system, workload, control-plane, storage, Pod, and GPU monitoring components, Managed Prometheus, and Dataplane V2 observability metrics; uses gcloud logging read to query system, workload, and audit logs; uses kubectl top to inspect live Pod, container, and node resource usage; and provides metric, LQL, alert-threshold, node-health, and cost-reduction examples.
- A GKE platform engineer needs to enable a production monitoring suite that includes API server, Scheduler, and Controller Manager metrics.
- An operations engineer needs to query Cloud Logging entries for a specific cluster, namespace, or container.
- An SRE needs Managed Prometheus metrics for Pods, nodes, the control plane, or GPUs.
- A troubleshooter needs to investigate OOMKilled events, scheduling failures, API latency, or nodes that are not ready.
- A team needs to reduce non-production monitoring to system-only metrics to control costs.
How do you install this skill?
- Before running cluster-update commands, confirm scope, permissions, cost impact, and rollback; --quiet is not a safety confirmation.
- Verify monitoring parameter names against the target GKE version and region, especially the terminology differences between the tables and CLI examples.
- Logs and audit logs may contain sensitive data; retention, access control, redaction, and data-residency guidance are missing.
- For mainland-China users, verify network reachability of Google Cloud, gcloud, monitoring consoles, and related APIs.
- Shell / CLI
- Network access
- MCP Server
gcloud CLIkubectlGKE MCP tools
Install the repository in an Agent Skills-compatible client with: npx skills add google/skills. Select gke-observability during installation; its source file is skills/cloud/gke-observability/SKILL.md. The repository contains 90 separate skills, and the installation command supports selecting individual skills.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/google/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/cloud/gke-observability" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Configure complete observability for my GKE cluster, including Cloud Logging, Cloud Monitoring, Managed Prometheus, and control-plane metrics.
After installation, use a prompt such as “Configure complete observability for my GKE cluster, including Cloud Logging, Cloud Monitoring, Managed Prometheus, and control-plane metrics.” You can also run the documented gcloud, kubectl, and gcloud logging read commands; cluster and Kubernetes resource diagnostics can use the listed GKE MCP tools. The source does not specify client-specific trigger syntax, permission setup, or retry behavior.
What are this skill's strengths and limitations?
- Covers GKE logging, monitoring, metrics, audit logs, alerting, and cost considerations.
- Explicitly includes API server, Scheduler, and Controller Manager control-plane metrics in its golden path.
- Documents the boundaries and examples for MCP, gcloud, and kubectl operations.
- Includes Managed Prometheus, GPU, node-health, and common troubleshooting metrics.
- Focused on GKE and Google Cloud; it does not cover local logging frameworks or external APMs outside GKE.
- Full monitoring and logging may incur Cloud Logging, custom Cloud Monitoring metric, and Managed Prometheus charges.
- The source provides no test suite, version compatibility matrix, or permission requirements.
- Cloud Trace and Cloud Profiler are production recommendations, not golden-path defaults.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| GKE Observability Configuration this page ✓ Google · Official | 44 · Not recommended | ★ 21k | 12d ago | Apache-2.0 |
| GKE Cluster Autoscaler Guide ✓ Google · Official | 52 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| GKE Production Readiness Review ✓ Google · Official | 49 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| GKE Production Golden Path ✓ Google · Official | 54 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| GKE GPU/TPU Disruption Diagnosis ✓ Google · Official | 50 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill explicitly covers logs, audit logs, metrics, and cluster updates, includes cost warnings, and limits scope to GKE, so no obvious overreach or covert behavior is shown. Points are deducted because mutating commands use --quiet without requiring confirmation, and the skill omits required permissions, sensitive-log handling, data-flow disclosure, rollback, and pre-change recovery guidance; comprehensive telemetry defaults may also increase cost and exposure.
The document covers a golden path, queries, cost reduction, and several diagnostic cases, making the happy path plausible. Points are deducted for the absence of tests, input validation, failure handling, or diagnostic feedback; configuration terminology may be inconsistent with CLI terminology (for example APISERVER/API_SERVER and SYSTEM_COMPONENTS/SYSTEM), and MCP tools are listed without parameters or invocation workflows. Static review cannot establish reliability above the medium range.
Trigger conditions and the non-fit boundary are stated clearly, covering GKE logging, monitoring, Prometheus, kubectl, and gcloud use cases. Points are deducted because inputs, output format, permission prerequisites, and cluster/version boundaries are not defined, Chinese-language support is absent, and core operation depends on Google Cloud, GKE, gcloud, kubectl, and overseas cloud services whose mainland-China reachability is not addressed.
Repository context supplies information architecture, naming, installation, Apache-2.0 licensing, support channels, and an active-development note; the skill includes sections, examples, metrics, cost guidance, and some limitations. Points are deducted because the skill lacks a version, changelog, named maintenance owner, or explicit update path, dependency/version requirements are incomplete, MCP-versus-CLI usage is underspecified, and some Markdown tables are malformed.
The skill provides fairly direct commands and recommendations for GKE observability configuration, log queries, live usage, alerting, and metric selection, so it can address the core task. Points are deducted because commands rely on undefined placeholders, lack preflight checks, result verification, recovery, and directly deliverable configuration output; the broad default may require substantial manual cost and suitability review. Static calibration caps this dimension at 7.
Specific CLI commands, metric names, query expressions, and Google documentation links provide some auditability. Points are deducted because there are no committed tests, CI evidence, execution results, version pinning, or third-party reproduction records; the links were not independently verified in this static review, so the score cannot be higher.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →