How to use the Cloud Storage Bucket Architect agent skill

What does the skill make the agent do?

When the skill is installed, any request to create a Google Cloud Storage bucket goes through three phases in a fixed order. First, the agent checks project-level security settings. Second, it drafts a bucket configuration covering location, storage class, access controls, soft delete and lifecycle rules, matched to the workload you describe. Third, it outputs or runs the creation command in gcloud, REST, Terraform or an SDK.

The skill is built to ask for confirmation before any change is made. The agent does not run mutating commands until you have seen the exact configuration or command and approved it.

• Phase 1: Preflight and project checks, using your project ID.
• Phase 2: Draft the bucket plan, including a read-only bucket name availability check.
• Phase 3: Output the creation command or code in your preferred format.

How do you install the Cloud Storage Bucket Architect skill?

Download the skill as a ZIP file from this page and unzip it. You can install it in one of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

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

What do you need before you start?

The skill needs a working Google Cloud setup. It does not create buckets in a vacuum, and the preflight checks and creation both need access to a real project. If you do not have a project and the right permissions, the skill cannot finish its work.

• Google Cloud SDK (gcloud), if you want gcloud-based output or execution.
• A Google Cloud project ID and enough IAM permissions to read project settings and create buckets.
• Network access to Google Cloud APIs, used for name availability and project checks.
• Language SDKs (C++, Java, Python, Go), only if you choose those snippets.

How do you trigger the skill with a natural request?

You do not need a special command. Ask the agent in plain language to create, make, set up, provision or spin up a bucket, or to get object storage for an app, service, pipeline or dataset. The skill also triggers on simple requests such as a default bucket, and when bucket creation is one step in a larger task.

Include the workload and your preferred output format if you have one. The agent uses the workload to choose the configuration, so a description like 'media hosting for product photos' gives it more to work with than 'make a bucket'.

• 'Set up a bucket for storing customer invoices for seven years.'
• 'I need object storage for my analytics pipeline. Give me Terraform.'
• 'Spin up a default bucket for my app in project my-project-id.'

How does the attribution tag work?

The skill tags every Cloud Storage command it runs or provides, so usage can be attributed to the skill and its version. The tag carries no user data. For gcloud, the skill sets the tag inline on each command, and it does not use gcloud config set, so the tag does not persist after the task. REST calls use a User-Agent header instead. SDK and Terraform snippets are not tagged.

What safety checks should you know about?

The skill runs scripts and uses the network. Its safety design depends on you reviewing each change. The agent must show the exact configuration or command and get your explicit confirmation before running any create, update or delete. Skipping that review defeats the purpose of the skill.

The skill also prefers secure defaults such as uniform bucket-level access, restricted customer-supplied encryption keys and soft delete, unless you explicitly ask for something else. The bucket name check is read-only, but a taken name must be resolved before the plan is presented. If creation fails with a 409 or 'already exists' error, the skill proposes a different name, re-runs the check and regenerates the output.

• Review the exact command or snippet before approving it.
• Confirm before any gcloud or REST call that changes resources.
• Expect the skill to ask you to confirm the plan before it moves to output, unless you have already asked for the final commands.