What does an AI agent need before it can make a useful change to your project? In this guided session, Keith Hutchinson walks through a practical way to give an agent context, a clear task, and boundaries for its work.

Using the M.A.S.C.O.T. framework, we’ll build a project briefing in AGENTS.md and a focused task specification in spec.md. Along the way, we’ll look at observable acceptance criteria, review evidence, and decisions that should come back to a human.

You’ll leave with a repeatable approach you can adapt to your next small change: brief the agent, define the work, review its approach, and check the result.

40-minute guided presentation for a mixed technical and nontechnical audience. No setup or coding required from attendees. Includes six minutes of presenter-led construction and three minutes of critique, with prepared examples for a quiet room.