AI agents promise to write your code, ship your features, and close your tickets before standup. In practice? They hallucinate files that don't exist, refactor things you never asked them to touch, and confidently break your build while telling you everything looks great.

It's time for an intervention.

In this workshop, we'll explore the full spectrum of agentic software development. From overly cautious agents that won't act without hand-holding to reckless ones that rewrite your entire architecture on a whim.
You'll learn where the levers are and how to pull them without everything catching fire. We start at the foundation: context engineering. What an agent knows determines what it does, and what it forgets (context rot) determines what it destroys. Get this wrong and nothing else matters.

Once the agent sees clearly, you need it to behave. That's where skills, instructions, and hooks come in. Repeatable workflows, hard boundaries, and audit points that let you intercept before damage is done.

Then we go structural. The Agent OS workflow introduces spec-driven development: structured feature specs that turn vague intent into verifiable, shippable work. No more hoping the agent guesses right.

By the end, you'll be crafting custom commands and skills tailored to your codebase, your conventions, your way of working, until the agent does exactly what you tell it. Every time. You'll leave with working code, battle-tested configurations, and the confidence that when you say "ship it," it actually ships what you meant.

Who should attend: Developers and technical leads who've used AI coding agents in real projects and are tired of babysitting them. Bring a laptop.