Google Agents CLI Workflow (ADK Lifecycle)

This skill manages the full lifecycle of building an AI agent from scratch with Google's Agent Development Kit (ADK) and the google-agents-cli tool: understand → review samples → scaffold → code → evaluate → deploy → publish → observe. It works with any coding agent such as Gemini CLI, Claude Code, or Codex.

When to use it? For agent build, evaluation, and deploy needs on Google Cloud, such as "develop an agent with ADK", "run the agent locally", "test/deploy the agent", or "publish to Agent Runtime". When targeting the Vertex AI / Gemini Enterprise Agent Platform, it guides you to the correct CLI flags and phase ordering.

What it produces? A phase-by-phase development plan and executable command blocks: scaffold create/enhance, smoke testing with agents-cli run, behavior validation with eval run, deploy, and observability setup. It also applies code protection (no changing model/version), the use of eval instead of pytest, product name mapping (Vertex AI → agent_runtime), and systematic debugging rules.

The output is a Turkish guide report enriched with real bash command blocks. Deploy always requires explicit human approval; the model is not changed unless explicitly requested.