Skip to content
Get packs

Guide · Agent skill

How to use the GKE Observability agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The gke-observability skill is a set of instructions that teaches an AI agent how to configure and troubleshoot monitoring and logging on Google Kubernetes Engine (GKE). It is written for platform and DevOps engineers who run GKE clusters and want a consistent monitoring setup. After this guide you will know what the skill does, how to install it, how to phrase requests so the agent uses it, and which safety points to check before letting it change a cluster.

GKE Observabilitygoogle/skills · Apache-2.0

Quick start

  1. 1Download the ZIP. Download the gke-observability ZIP file from this page.
  2. 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for all projects, or into .claude/skills/ inside one project.
  3. 3Or upload it in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
  4. 4Prepare gcloud and permissions. Make sure the Google Cloud SDK is installed and you have access to a project with GKE and monitoring permissions.
  5. 5Ask with a specific request. Describe the setup or the symptom, such as missing metrics or unhealthy scrape targets, and mention GKE.
  6. 6Review before applying. Check the commands and component changes the agent proposes, especially on production clusters.

What does the gke-observability skill make the agent do?

Once installed, the agent follows Google's recommended setup for monitoring and logging on GKE. The skill covers Cloud Logging, Cloud Monitoring and Google-managed Prometheus (Managed Service for Prometheus, GMP).

It can also help diagnose common problems such as missing metrics, unhealthy scrape targets, PodMonitoring misconfigurations, failing alert rules and permission errors. The skill also covers control-plane metrics for the API server, scheduler and controller manager.

  • Logging: system components and workloads are sent to Cloud Logging by default in the golden path.
  • Monitoring: a broad set of components is enabled, including control-plane components, kubelet, cAdvisor and DCGM.
  • Managed Prometheus: enabled in the golden path, with Dataplane V2 flow metrics turned on.
  • Troubleshooting: missing metrics, scrape targets, PodMonitoring, alert rules and permissions.

Install the gke-observability skill

Download the ZIP file for the skill from this page. Then choose one of the two install methods below.

For a personal setup across all your projects, unzip the file into ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project. In both cases the skill folder should contain SKILL.md and references/attribution.md.

In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.

Give the agent the right kind of request

The agent loads the skill when your request matches its description. Name the thing you want to set up or fix, and mention GKE so the agent does not pick a generic answer.

Good triggers include configuring GKE monitoring, setting up GKE logging, configuring Prometheus metrics collection, and troubleshooting Managed Service for Prometheus issues. Requests about local application logging frameworks or external APM tools outside GKE do not match this skill.

  • Set up Cloud Logging and Cloud Monitoring for my GKE cluster.
  • Why are no metrics showing up from my PodMonitoring resource in GKE?
  • My Managed Prometheus scrape targets are unhealthy. Help me find out why.
  • Check whether my alert rules are failing to evaluate on this GKE cluster.
  • Enable control-plane metrics for the API server and scheduler on my cluster.

What the skill needs before it runs

The skill expects the Google Cloud SDK (gcloud) for terminal commands. You also need access to a Google Cloud project with GKE and monitoring permissions. Optionally, MCP tools for GKE such as get_cluster and get_k8s_logs can be used.

Monitoring changes are made with gcloud or API commands. For example, control-plane metrics are enabled with flags on gcloud container clusters update.

  • Google Cloud SDK (gcloud) for terminal commands.
  • A Google Cloud project with GKE and monitoring permissions.
  • Optional: MCP tools for GKE, such as get_cluster and get_k8s_logs.
gcloud container clusters update --monitoring=SYSTEM,API_SERVER,... --enable-managed-prometheus

Safety notes: scripts, network and cost

This skill does not run scripts, but it does use the network, because it calls Google Cloud services through gcloud, REST APIs or MCP tools.

Before any gcloud, REST API or MCP command, the agent is instructed to read the attribution reference and tag the command with a usage-attribution label. For terminal commands the label is set inline on each call, not through gcloud config set, so it does not persist.

The golden path enables many components that may add cost and load to your cluster. Review the change before applying it on clusters where monitoring changes are risky.

  • Do not use the skill for non-GKE Kubernetes or other clouds; the Google-specific settings will not apply.
  • Check cost and cluster load before accepting the full golden path component list.
  • Review generated changes before they run on production clusters.
CLOUDSDK_METRICS_REQUEST_ATTRIBUTION="agent-skills/1.0 (skill:gke-observability)" \ gcloud <command> [flags]

Tips for getting good results

Name your cluster, project and symptom in the request. The agent can work faster when it knows whether you are setting up monitoring for the first time or debugging an existing failure.

When troubleshooting missing metrics, ask the agent to check scrape targets and PodMonitoring configuration together, since the skill covers both. For log questions, the gcloud logging read command and the get_k8s_logs MCP tool are the relevant tools.

  • State whether you want setup or troubleshooting.
  • Give the cluster name and project ID.
  • Ask for logs with gcloud logging read when you need raw log entries.

What people use it for

  • Setting up Cloud Logging, Cloud Monitoring and Managed Prometheus consistently across GKE clusters.
  • Finding out why a PodMonitoring resource is not producing metrics or why scrape targets are unhealthy.
  • Enabling control-plane metrics for the API server, scheduler and controller manager to diagnose slow API responses or scheduling delays.

Frequently asked questions

How do I install the gke-observability skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. In the Claude apps you can upload the ZIP under Settings → Capabilities → Skills.

What do I need to use the gke-observability skill?

You need the Google Cloud SDK (gcloud) for terminal commands and access to a Google Cloud project with GKE and monitoring permissions. MCP tools for GKE, such as get_cluster and get_k8s_logs, are optional.

Is the gke-observability skill free?

The skill is licensed under Apache-2.0. This page does not list pricing for Google Cloud services, and the monitoring components the golden path enables may add cost to your cluster.

Does this skill work with clusters that are not on GKE?

No. The skill is built around Google-specific settings for GKE. It is not intended for non-GKE Kubernetes or other clouds, where those settings will not apply.

Can the skill help with application logging or external APM tools?

No. The skill explicitly excludes configuring local application logging frameworks and external APM tools outside GKE.

Does the skill run scripts or use the network?

It does not run scripts, but it uses the network because it calls Google Cloud services through gcloud, REST APIs or MCP tools.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you run GKE clusters and want a consistent, Google-recommended monitoring and logging setup, or need help fixing Managed Prometheus collection problems. Pick something else if you only need application logging inside your code, external APM outside GKE, or non-GKE Kubernetes, and review the golden path's added cost and load before applying it to sensitive production clusters.

Download the skill