Building an AI agent often starts with a prompt, a model, and a few tools. But without a clear plan, the agent can quickly become difficult to test, maintain, and deploy.

In this session, we will use a spec-driven approach to build an AI agent with Google ADK, Antigravity, and Agents CLI. We will define what the agent should do, identify the tools it needs, and describe its expected behavior.

We will then use Antigravity and Agents CLI to turn the specification into an implementation plan and scaffold the ADK application. Finally, we will explore the core ADK components used to build the agent, including instructions, tools, sessions, state, and workflows.