In this workshop we get hands-on with Gemini CLI, MCP and agent skills.

We start with a demo of Gemini CLI, showing how we can combine features like Google Search, shell execution and MCP servers to perform sophisticated tasks. We cover configuration, including model selection and the policy engine.

You will learn how to automate interactions with a web application in the browser, using natural language prompting. There are many possibilities for what we can do with this, and in this session we will experiment with automated UI testing.

We will run through a dedicated Google codelab, step-by-step. We start with MCP-based remote control, and then move on to using Agent Skills. You will learn what agent skills are, and the scenarios where we can combine them with MCP or a CLI. We cover the advantages of Agent Skills, including just-in-time power-ups through progressive discovery. Think: Neo learning kung-fu!