What does this skill do, and when should you use it?
This skill answers GKE cost questions using billing exports, cost-allocation metadata, and live cluster metrics. It can break down historical costs by project, cluster, namespace, and workload, while also checking budgets and current resource consumption. Its analysis distinguishes Autopilot pod-request billing from Standard node-VM and control-plane charges. It is intended for diagnosis, not for applying optimization changes or creating rightsizing configurations.
Reads cost and credits data from the GCP Billing Detailed BigQuery Export and uses bq templates to query the last 30 days of workload, cluster, and namespace costs; checks GKE Cost Allocation labels; lists billing budgets with gcloud billing budgets list; compares live node and pod usage with kubectl top nodes and kubectl top pods --all-namespaces --containers; and can provide the gcloud command for enabling cost allocation.
- A platform engineer needs to compare GKE costs across projects and aggregate them by workload.
- A FinOps analyst wants to identify the most expensive namespace in a cluster using BigQuery cost-allocation labels.
- A GKE operator is investigating a cost spike and needs to compare pod requests with actual CPU and memory usage.
- A Standard-mode cluster owner wants to assess idle nodes, low-utilization development clusters, and control-plane fees.
- A team needs to understand how CUDs or Spot VMs affect gross and net billing amounts.
How do you install this skill?
- The “read-only CLI commands” description is inaccurate: gcloud container clusters update --enable-cost-allocation mutates cluster configuration and should require explicit confirmation plus rollback guidance.
- The BigQuery queries depend on correct table paths, export schema, labels, and credit semantics; users should verify export enablement, permissions, and accounting methodology before relying on results.
- kubectl top requires working metrics availability and cluster access; missing metrics should produce a diagnostic path rather than being assumed available.
- Chinese-language support and mainland-China reachability are unspecified; core Google Cloud, BigQuery, and GKE API dependencies may be affected by the network environment.
- Shell / CLI
- Network access
BigQuery billing detailed exportBigQuery CLI (bq)Google Cloud CLI (gcloud)kubectlGKE cluster
Run npx skills add google/skills, then select the skills to install from the repository. The source does not document the installation directory or a separate command for installing only gke-cost-analysis.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/google/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/cloud/gke-cost-analysis" ~/.claude/skills/
rm -rf "$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:
- Which namespace had the highest GKE cost in the last 30 days?
- Why is this GKE cluster cost increasing?
After installation, ask a concrete question such as “Which namespace had the highest GKE cost in the last 30 days?” or “Why is this GKE cluster cost increasing?” Provide the full BigQuery billing-export table path, including the billing account ID, and supply project, region, cluster, namespace, or workload values when relevant.
What are this skill's strengths and limitations?
- Supports cost analysis at project, cluster, namespace, and workload granularity.
- Combines historical BigQuery billing data with live `kubectl` utilization data.
- Provides copyable `bq`, `gcloud`, and `kubectl` command templates.
- Explains distinct Autopilot and Standard cost drivers.
- Requires a configured GCP Billing Detailed BigQuery Export.
- Namespace, label, and workload analysis requires GKE Cost Allocation to be enabled.
- The source provides no test suite, platform test results, or detailed failure-handling evidence.
- It does not apply optimization changes, create VPA/MPA rightsizing manifests, or select ComputeClasses.
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 Cost Intelligence this page ✓ Google · Official | 50 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| GKE Cost Optimization ✓ Google · Official | 47 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| GKE Basics Navigator ✓ Google · Official | 45 · Not recommended | ★ 21k | 12d ago | Apache-2.0 |
| GKE Backup & Disaster Recovery ✓ Google · Official | 35 · Not recommended | ★ 21k | 12d ago | Apache-2.0 |
| GKE ComputeClasses Advisor ✓ Google · Official | 57 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
The source explicitly delegates optimization changes to the gke-cost-optimization skill; it provides no other verifiable competitor or alternative comparison.
How did FollowSkills review this skill?
The skill identifies BigQuery, gcloud, and kubectl as data and control surfaces, and explains the billing and cluster-metric sources. However, it does not specify required IAM permissions, sensitive billing-data handling, credential boundaries, or exfiltration risks. It also calls the commands read-only while including a mutating gcloud container clusters update command, without user confirmation, rollback, or pre-change checks, so points are deducted.
The guidance, query templates, defaults, and main analysis paths are reasonably coherent. Reliability is reduced because the read-only claim conflicts with the cost-allocation update command, while availability of kubectl top, billing exports, permissions, empty-result behavior, and abnormal-input diagnostics are not addressed. Static evidence is insufficient for a higher score.
The audience, GKE cost-analysis scenarios, input placeholders, and out-of-scope optimization work are fairly clear, with an Autopilot/Standard distinction. Trigger precision and boundaries for non-GKE costs or missing exports and metrics remain incomplete. Chinese-language interaction and mainland-China reachability are not addressed, which may limit use if the core services are inaccessible.
The document has clear layering and includes an overview, key points, commands, query examples, prerequisites, Apache-2.0 licensing, installation guidance, and an issue path. It lacks skill-level versioning, changelog, explicit maintenance ownership, complete dependency and permission notes, and troubleshooting or FAQ coverage.
The skill supplies reusable BigQuery queries and practical methods for cost, label, and utilization analysis, covering the core task. However, every parameter requires manual replacement, the file contains no execution evidence, and users must still validate permissions, export schema, credit semantics, and metric availability, so the result is not shown to be directly usable without review.
The source provides auditable query templates, commands, stated prerequisites, licensing, and repository maintenance signals. It provides no skill-specific test suite, CI execution evidence, third-party corroboration, or reproducible run results. Static review therefore supports only limited verifiability.
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 →