How to use the Google Agents CLI Onboarding agent skill

What does the skill make the agent do?

When installed, the skill lets your agent walk you through a one-time setup that installs the agents-cli CLI and related specialized skills. After that, it tells the agent which skill to load for each phase: defining the agent, scaffolding a project, writing code, evaluating, deploying to Google Cloud, publishing to Gemini Enterprise, and monitoring.

Before any code is written, the skill asks for your requirements (agent purpose, external tools, deployment target) and records them in a spec file named .agents-cli-spec.md.

• Phase 0 Understand and Phase 1 Study use google-agents-cli-workflow
• Phase 2 Scaffold uses google-agents-cli-scaffold
• Phase 3 Build uses google-agents-cli-adk-code
• Phase 4 Evaluate uses google-agents-cli-eval
• Phase 5 Deploy uses google-agents-cli-deploy
• Phase 6 Publish uses google-agents-cli-publish
• Phase 7 Observe uses google-agents-cli-observability

Install the skill from the ZIP

Download the ZIP for this skill from this page and unzip it. Put the resulting folder in your personal skills directory, ~/.claude/skills/, or in a project's .claude/skills/ directory if you want the skill to apply only to that project.

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

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

Run the one-time setup

Once the skill is loaded, run the setup command from the skill's tip box. It installs the CLI and enables all 7 specialized development skills in your coding agent. If you only want the expert skills and prefer the agent to handle execution, the source lists an alternative command.

After setup, ask the agent to run agents-cli info to confirm the installation and see the active project configuration.

Trigger the skill with natural requests

You do not need a special command to use the skill. It activates when your request matches its description, such as asking to create a new agent, develop an agent using ADK, run the agent locally, debug agent code, test, evaluate, deploy, publish or monitor an agent. The skill also activates when you need the ADK (Agent Development Kit) development lifecycle.

Describe what you want in plain language. The agent will then load the specialized skill that matches the current phase.

• "Create a new agent using ADK that answers questions about our product docs."
• "Evaluate my agent and grade its traces."
• "Deploy this agent to Cloud Run."
• "Publish my deployed agent to Gemini Enterprise."

Know what you need before you start

The skill needs uv (for the uvx command) or Node.js with npx. It also uses the google-agents-cli package from PyPI or the google/agents-cli repository. Most later steps deploy to Google Cloud, so you also need the gcloud CLI, a Google Cloud account and project, and for the publish step, Gemini Enterprise access.

The skill is only a routing entrypoint. Without a Google Cloud setup, the deployment and publishing steps will not work.

• uv or Node.js/npx
• google-agents-cli from PyPI or google/agents-cli on GitHub
• gcloud CLI and a Google Cloud account/project for deployment and publishing
• Gemini Enterprise access for the publish step

Safety notes: scripts and network access

This skill runs scripts and uses the network. The setup command downloads and executes third-party tooling, so it runs external commands with network access. If you want to review or audit code before it runs, read the skill files and the setup source first, and do not run the setup command until you are comfortable with it.

The skill also asks the agent to tag gcloud commands and direct Google Cloud API or MCP requests with a usage attribution value. Per the attribution file, this tag identifies only the skill and its version and carries no user data.

• The setup command downloads and runs external tooling
• Review the files (SKILL.md and references/attribution.md) before use
• gcloud commands run with an inline attribution environment variable, not a persistent gcloud config change

Tips for a smooth start

Give the agent your requirements up front so the spec file is complete before coding. Confirm each phase before moving on, since the skill is organized as a sequence. If something breaks, check the agents-cli --help output for the full list of commands and global options.

Report bugs or improvements at the Google Agents CLI issues page on GitHub. The Google Agents CLI documentation is also linked from the skill.

• Name the agent purpose, tools and deployment target in your first message
• Keep the .agents-cli-spec.md file in your project as the single source of requirements
• Run agents-cli --help for the full command list