Session Recap
What we covered and how to apply it this week.
A brief is not a prompt. It is a permanent job description for an agent that will run dozens of times. The four components of a strong brief are context (who the agent is), task (what it does), constraints (what it never does), and output format (exactly what success looks like). Most agents fail not because the AI is bad, but because the brief is missing one of these pieces. When you write all four, your agent works. When you skip even one, you get inconsistency and wasted time.
Take one agent you built in the past month. Look at the instructions you gave it. If you cannot point to a clear output format section, rewrite that part now using a template or real example. Run it on one real input and see if the output matches your template. You will probably see something you need to fix.
Build a production-quality brief for an agent using the 4-component framework. Test it on real data. Submit your brief and your test result to your AI Buddy for grading. The grading criteria are simple: completeness, output format specificity, edge case coverage, and clarity. If someone else could follow your brief on their first day without a follow-up conversation, you have done it right.
Go to challenge →