---

name: cli-configuration-deployment description: "Use ADK CLI commands, agent/app discovery, YAML configuration, local run/web/API server flows, eval/test commands, deployment commands, and safe CLI/config inspection." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

ADK CLI, Configuration, and Deployment

Use this sub-skill when a user asks how to run or expose an ADK app with the adk console script, why an agent is not discovered by adk web, how to structure a Python or YAML agent app, how to validate generated YAML config, or how to reason about adk deploy targets and credentials. ADK 2.3.0 provides the google.adk import root and the adk console command on Python 3.10+.

Route Common Requests

• Run locally: Use CLI reference for adk run, adk web, adk api_server, adk test, adk eval, adk eval_set, and safe --help checks.
• Fix app discovery: Use configuration for loader conventions, root_agent/app, agent.py, __init__.py, root_agent.yaml, .env, and service registration.
• Generate or validate YAML: Use configuration for AgentConfig.json, agent_class, sub_agents, tools, callbacks, CodeConfig, and schema-validation workflow.
• Deploy safely: Use CLI reference for adk deploy cloud_run, adk deploy agent_engine, and adk deploy gke; use troubleshooting for credentials, project/region, optional extras, and generated source caveats.
• Diagnose failures: Use troubleshooting for missing agents in web UI, server/log/CORS issues, YAML schema errors, deployment failures, missing optional dependencies, and model credential problems.
• Inspect an installation: Run inspect_adk_cli.py to list available CLI commands and locate installed config schema metadata without starting servers, deployments, network calls, or credential flows.

Boundaries

• Included here: adk console commands; agent loader and app discovery; root_agent/app; YAML config and schema metadata; local run, web, api_server, test, eval, and eval_set; deploy command families; service registry; graph visualization; safe CLI help diagnostics.
• Route elsewhere: Eval assertion design and detailed metrics go to evaluation-debugging; Python Agent/LlmAgent object design goes to agent-construction; repository maintainer commands, schema regeneration PR workflow, and source test policy go to repo-development.

Safe Defaults

• Prefer adk --help and subcommand --help for non-mutating CLI checks.
• Do not launch adk web or adk api_server unless the user explicitly wants a local server.
• Do not run adk deploy ... without confirmed project/region/credentials and a clear deployment target.
• Treat base installs as intentionally minimal: optional DB, extensions, MCP, eval, GCP/cloud, and A2A integrations can be absent and should be diagnosed rather than assumed.