How to use the Google Cloud Performance 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 loads when a task matches it. This skill is built from the Performance Optimization pillar of Google Cloud's Well-Architected Framework, a set of best-practice guidelines. When it is active, the agent can evaluate a Google Cloud workload against that pillar, ask structured questions about resource sizing, caching, modular design and monitoring, and give recommendations on compute, storage, scaling and observability.

The skill also includes a validation checklist you can use to judge how well an existing architecture follows these practices. It is guidance and questions, not tooling: it does not change resources or make a live environment faster by itself.

• Covers four principles: plan resource allocation, take advantage of elasticity, promote modular design, and continuously monitor and improve performance.
• Names example Google Cloud products relevant to each, such as Compute Engine MIGs, GKE, Cloud Run, Cloud CDN, Memorystore, Bigtable, Spanner, Cloud Trace, Cloud Profiler and Cloud Monitoring.
• Provides a checklist covering resource allocation, modular design, monitoring and continuous improvement, and elasticity.

Install the skill from the ZIP file

Download the ZIP file for this skill from this page, then unzip it. The skill is a folder that contains SKILL.md and a references folder with attribution.md. Place the folder in one of these locations: ~/.claude/skills/ to make it available to you across projects, or a project's .claude/skills/ folder to make it available only in that project.

If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills rather than copying files by hand.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into the .claude/skills/ folder of your project
• Claude apps: Settings → Capabilities → Skills, then upload

How to trigger the skill with natural requests

You do not need a special command. Describe the task in plain language and mention performance, Google Cloud or the Well-Architected Framework. The skill's description says it applies when you want to evaluate a workload, identify performance requirements, or get recommendations on resource allocation, modular design and elasticity.

Give the agent as much context as you can about the workload: the Google Cloud products involved, expected traffic, and what you already monitor. The more specific your request, the more useful the questions and recommendations will be.

• "Review the performance design of my Cloud Run and Cloud SQL app before launch."
• "Use the Well-Architected Framework performance pillar to check our caching strategy."
• "Give me structured performance questions for tomorrow's architecture review."
• "Check this architecture against the performance checklist and tell me which items are missing."

What the skill asks and what you get back

The agent asks questions drawn from the skill's assessment list, such as how you size compute for expected peak loads, which caching layers you use, which architectural patterns you employ, how often you review production performance, and how you handle sudden traffic spikes. Answering them lets the agent tailor its recommendations.

Recommendations are grounded in the four principles, and the skill points to a grounding document for each. The checklist lets you mark items against your architecture, for example whether auto-scaling rules are configured and validated, or whether circuit breakers or bulkheads isolate failures.

Safety notes: scripts and network access

The skill runs no scripts. It does use the network: its principles reference documentation URLs on docs.cloud.google.com, and its attribution reference covers any gcloud, REST API or MCP commands the agent may run. Before running any gcloud, REST or MCP command with this skill, the agent is instructed to follow the attribution file. That file asks the agent to prefix gcloud commands with a metrics environment variable set inline, and not to use gcloud config set for it. It states the tag identifies only the skill and its version and carries no user data.

The skill lists no required tools or accounts. Still, review any command the agent proposes before you let it run against a live project.

• Runs scripts: no
• Uses network: yes
• Requires: nothing listed

Tips for a useful review

Treat the output as a structured starting point for a design review, not a verdict. Bring real numbers where you have them, such as load test results or historical traffic, because the checklist asks whether provisioning is based on load testing or historical data rather than general estimates.

The skill is specific to Google Cloud products, so it is not suited to workloads on other cloud providers.

• Bring load test results or historical traffic data to the review.
• Ask the agent to work through one principle at a time for a deeper answer.
• Use the checklist as a record: note which items you meet and which you do not.