How to use the Google Cloud Storage Basics agent skill

What is an agent skill?

An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches it. This skill is published under the Apache-2.0 licence and its main instructions live in a file called SKILL.md, with supporting reference files alongside it.

What the skill makes the agent do

The skill covers creating buckets, uploading, downloading, copying, moving and deleting objects, setting access controls, lifecycle rules, versioning and storage classes. It also includes reference material on client libraries, Terraform, gcsfuse mounts, transfers and performance tuning.

If a Cloud Storage MCP server is connected, the agent calls its structured tools. Otherwise it uses the gcloud command line or the API. The skill explicitly excludes block storage (Persistent Disk), BigQuery and database work.

Install the skill

You can install the skill in one of two ways. For a personal setup, download the ZIP from the skill page and unzip it into ~/.claude/skills/. To scope it to one project, unzip it into that project's .claude/skills/ folder.

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

• Personal: ~/.claude/skills/
• Project: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills

Requirements before you start

You need a Google Cloud project and credentials with Cloud Storage permissions. The gcloud CLI is needed only for the fallback path; it is not needed when MCP tools are connected.

Optional extras are a local MCP Toolbox or the Google-hosted remote Cloud Storage MCP server, and Terraform, client libraries or gcsfuse for those specific tasks.

• Google Cloud project with Cloud Storage permissions
• gcloud CLI, for the fallback path only
• Optional: local MCP Toolbox or remote Cloud Storage MCP server
• Optional: Terraform, client libraries or gcsfuse for those tasks

Trigger the skill with natural requests

You do not need a special command. Describe the storage task in plain language, and name Cloud Storage or GCS so the agent loads the right skill. Good requests name the bucket, the object and the action.

The skill tells the agent to stop and ask for confirmation before deleting or overwriting anything, so expect a confirmation prompt for those actions.

• Create a bucket called my-reports-bucket in us-central1 and set it to Standard storage class.
• Upload ./exports/sales-2024.csv to gs://my-reports-bucket/exports/.
• List the objects in gs://my-reports-bucket/logs/ and show their sizes.
• Add a lifecycle rule to delete objects in gs://my-reports-bucket older than 90 days.

Safety notes: scripts and network

The skill runs scripts and uses the network, because it calls Google Cloud services with your credentials. Review what the agent plans to run before you approve it, and use a project and credentials you are comfortable with.

Deletes are hard to undo. The skill asks for confirmation before deleting or overwriting, but a mistaken delete in a production bucket can still cause data loss, so review bulk changes before approving them.

• Keep production buckets behind a review step for deletes and overwrites.
• Use least-privilege credentials that only grant the Cloud Storage permissions you need.
• Each MCP tool can be approved or blocked individually in your agent's settings.

Tips for better results

Name the bucket and full object path in every request so the agent does not guess. For production or workload-specific buckets, the skill routes to google-cloud-storage-bucket-architect before creating a bucket.

Use the MCP tools when you can. They run as single direct calls and do not need the gcloud CLI signed in. For code, the skill recommends client libraries with Application Default Credentials.

• Always include bucket and object names in full.
• Connect the MCP server if you run many interactive operations.
• Ask for client-library code or Terraform explicitly when you need it.