How to use the Google Cloud Cost Optimization Review agent skill

What is an agent skill and what does this one do?

An agent skill is a folder of instructions that an AI agent reads when your request matches the skill's description. This skill is built around SKILL.md, which tells the agent how to review Google Cloud workloads for cost efficiency and FinOps alignment.

When installed, the agent answers questions about Google Cloud spending with a structured checklist drawn from the Cost Optimization pillar. The checklist covers visibility (Cloud Billing reports, billing alerts and budgets, BigQuery billing export), rightsizing idle or oversized resources, and discount options such as Committed Use Discounts (CUDs), Sustained Use Discounts (SUDs) and Spot VMs. It also recommends Cloud Storage lifecycle policies and network choices that reduce transfer costs. Once the standard advice is given, the agent can ask follow-up questions about your workload to refine it.

• The skill is a set of instructions for the agent. It does not change any cloud resources by itself.
• It is scoped to cost only. It is not a general architecture design tool.
• It is not a price list. It is meant for reviewing spend, not looking up the current list price of one product.

Download and install the skill

Download the ZIP file from this page. Then choose one of the two install routes below.

For Claude Code, unzip the file into your personal skills folder at ~/.claude/skills/ so it is available in every project. To limit it to one project, unzip it into that project's .claude/skills/ folder instead.

In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/ (available across projects).
• Project install: .claude/skills/ inside your project folder.
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP.

Trigger the skill with natural requests

You do not need a special command. The agent loads the skill when your request matches its description, which covers analysing Google Cloud spend, reducing cloud bills, rightsizing resources such as Compute, GKE, Storage and Databases, evaluating discount options such as CUDs, SUDs and Spot VMs, and eliminating idle capacity.

Write the request as you would ask a colleague. Name the services and the cost problem. Adding the workload type and rough monthly spend helps the advice fit your situation.

• "Review our Google Cloud bill for the analytics project and tell me where to start."
• "Which of our Compute Engine instances look oversized, and how should we rightsize them?"
• "Should we buy committed use discounts for our steady GKE node pools, or is Spot better for the batch jobs?"
• "Our Cloud Storage costs keep rising. Help me plan lifecycle rules."
• "Find idle disks and old snapshots we are still paying for."

What you need before running live checks

The skill can answer from its checklist alone. If you want the agent to inspect your actual resources, you need a Google Cloud account with access to billing data for the workload, and the gcloud CLI, a REST API or MCP tooling that the agent can call.

The skill tells the agent to read its attribution file (references/attribution.md) before running any gcloud, REST API or MCP command. That file asks the agent to tag each command with a metrics environment variable that identifies only the skill and its version. The skill does not set this with gcloud config set.

• Google Cloud account with access to billing data for the analysed workload.
• gcloud CLI, REST API or MCP tooling, only if you want live commands.
• Read-only review first: the skill's own guidance is that it cannot change resources.

Safety notes: scripts and network use

This skill runs no scripts. Its files are SKILL.md and references/attribution.md, and it contains no executable code of its own.

It does use the network. The skill's guidance links to Google Cloud Well-Architected Framework documents as grounding sources, and any live check you enable will call Google Cloud services. Review each command the agent proposes before you approve it, especially anything that would modify, delete or purchase resources.

Do not treat the agent's advice as a binding financial decision. Before buying a CUD or deleting snapshots or disks, check actual usage data and your contract terms.

• No scripts are included.
• Network use: yes, for grounding references and any live gcloud, REST or MCP calls.
• Confirm deletions and purchases yourself; the skill does not perform them.

Tips for getting useful results

Start with visibility. The skill's guidance is to begin with Cloud Billing reports, filtered by labels such as department, and to use BigQuery billing export for custom analysis. Labels and Resource Manager structure make cost attribution easier, so mention them if your team uses them.

For storage, ask for Storage Insights analysis before lifecycle rules. The skill recommends checking current access patterns first. For databases, ask about high availability only for production, since HA can double node costs across environments.

Expect a full checklist first, then follow-up questions. The skill is designed to give the standard recommendations before asking about your workload, so answer those questions with real numbers where you can.

• Name the service (Compute Engine, GKE, Cloud Storage, a database) in your request.
• Mention your environments (production, staging, dev) so HA and sizing advice is specific.
• Ask for the Recommender and FinOps hub views to see existing savings opportunities.