Build Your Own AI Agent! From Idea to Running in 2 Hours

​We all have workflows we repeat every week. Writing briefs, summarizing research, prepping handoffs, reviewing work. Most of it could be partially automated, but setting that up always feels like a separate project.

​This session is about actually doing it, together.

​We'll start by mapping a real workflow each of us owns, identify what's worth turning into a skill or agent, then build and test it in the room. Along the way we'll look at what's already out there, including tools like Claude code, and compare how different systems handle memory, skill structure, and iteration.

​By the end, everyone should leave with at least one working skill or agent they built themselves.

• ​Mapping your workflow and identifying automation candidates

• ​Discussing when to use a skill vs. an agent (and why it matters)

• ​Reading through Claude's official documentation

• ​Benchmarking existing skills/agents

• ​Building and testing your own skill or agent live in Claude Code

• ​Investigating (1) File and markdown structure for maintainable agents (2)Workflow iteration loop: how to give feedback and improve outputs over time

• ​If we want, we can also explore Open Claw, Hermes Agent

​Format: 2 hours. Max 8 people. Laptops required.

​Who this is for: Designers,PMs/Developers.
Come with a workflow in mind if you can, but showing up without one is fine too.

​Requirement: Claude Pro subscription or above.