Google Agents CLI Agent Development Guide
Guides agents from specification through development, deployment, and monitoring.
The file makes the installation, development, deployment, publishing, and monitoring scope visible and requires the user to initiate the setup commands. However, it does not specify required permissions, data flows, sensitive-data handling, dependency supply-chain controls, external side effects, confirmation gates, or rollback, so points are deducted from the mid-level baseline.
The lifecycle table, command table, and next-step sequence are broadly coherent and include an installation-verification step. However, there are no tests, compatibility details, abnormal-input handling, or diagnosable failure feedback; the file also presents `uvx google-agents-cli setup`, `agents-cli`, and `npx skills add google/agents-cli` paths without in-file validation. The score is also constrained by static calibration.
The metadata clearly covers creating, developing, running locally, debugging, testing, evaluating, deploying, publishing, and monitoring agents. However, the trigger scope is broad, with no explicit non-fit cases, input/output boundaries, precise trigger rules, Chinese-language support, or mainland-China network reachability information, so points are deducted.
The document is readable and progressively organized by lifecycle phase, with installation commands, a command table, next steps, issue reporting, and supporting links. Repository context supplies an Apache-2.0 license and official-organization provenance. However, the skill lacks its own version, changelog, explicit maintenance owner, dependency versions, FAQ, and known-limitations section, so points are deducted.
As an entrypoint, it can guide a user from setup through the relevant development phases and provides copyable commands. However, its core capability depends on an external CLI and specialized skills outside the assessed path; command availability and final outputs are not verified in the file, leaving users potentially responsible for additional troubleshooting. The score is constrained by static calibration.
The file includes links to a GitHub issue tracker and documentation directory, and its workflow can be inspected. However, it contains no committed test suite, CI coverage, pinned dependency versions, or third-party execution evidence; key claims therefore remain primarily self-documented, warranting only limited verifiability.
- The setup commands introduce an external CLI and additional skills without specifying versions, permission scope, integrity checks, or uninstall/rollback procedures.
- Deployment, publishing, logging, and BigQuery analytics may create cloud resources, costs, and sensitive-data processing implications; the file provides no confirmation, least-privilege, or retention guidance.
- The trigger description is broad and does not address offline environments, Chinese interaction, or mainland-China network reachability.
What does this skill do, and when should you use it?
This is the entrypoint skill for Google’s agents-cli toolkit for building, evaluating, and deploying AI agents on the Gemini Enterprise Agent Platform. It routes users to specialized workflows for specification, scaffolding, ADK implementation, evaluation, deployment, publishing, and observability. After setup, it exposes commands such as agents-cli setup, scaffold, eval, deploy, and publish. The skill primarily initializes the environment, identifies the right lifecycle phase, and directs the next action.
Provides installation paths using uvx google-agents-cli setup or npx skills add google/agents-cli; guides users to run agents-cli info to verify installation; instructs them to capture the agent purpose, external tools, and deployment target in .agents-cli-spec.md; maps lifecycle phases to specialized skills; and documents CLI commands for scaffolding, evaluation, deployment, and Gemini Enterprise registration.
- A developer creating a new agent who needs to define requirements before generating a project.
- An engineer implementing agent logic, tools, callbacks, or state management with ADK.
- A team that needs to run, debug, or systematically evaluate an agent.
- A developer preparing deployment to Agent Runtime, Cloud Run, or GKE.
- A team publishing a deployed agent as a Gemini Enterprise tool and setting up monitoring.
What are this skill's strengths and limitations?
- Covers the agent lifecycle from requirements clarification through observation.
- Clearly separates workflow, scaffolding, ADK coding, evaluation, deployment, publishing, and observability phases.
- Includes copyable installation and core-operation commands.
- Connects workflows to Agent Runtime, Cloud Run, GKE, and Gemini Enterprise.
- It is an entrypoint and routing skill; implementation depends on the specialized skills it enables.
- The source provides no test suite, version requirements, or compatibility matrix.
- Detailed local-run and debugging commands are not documented here.
- Relevant Google Cloud configuration and permissions may be required, but no permissions checklist is provided.
How do you install this skill?
Place the skill folder in a directory supported by the Agent Skills-compatible client. To install the repository collection, run npx skills add google/skills and select the desired skill. To install the agents-cli expert skills, run npx skills add google/agents-cli, or run uvx google-agents-cli setup. The source does not specify the exact skill-directory path for each client.
How do you use this skill?
After installation, run agents-cli info to verify the setup. Ask the coding agent to create, develop, run, debug, test, evaluate, deploy, publish, or monitor an agent; begin Phase 0 by documenting the agent purpose, external tools, and deployment target in .agents-cli-spec.md. Use agents-cli scaffold <name> to create a project, agents-cli eval run to run and grade it, agents-cli deploy to deploy it, or agents-cli publish gemini-enterprise to register it.