Before you build: a design brief and your workspace

Design before you click

The most common reason a first agent disappoints is that the builder started typing instructions before deciding what the agent is for. Spend five minutes on a design brief. Every lab in this course starts from one.

A design brief answers six questions:

1. Goal. What outcome do I want, in one sentence?
2. Input. What does the agent start with? A web address, a question, a message body, nothing?
3. Tools. Which tools does it need, and what is the least it can have?
4. Output. What should the result look like, and where does it go?
5. Approvals. Which actions must wait for me?
6. Limits. What is a sensible budget per run, and how often should it run?

Check your workspace

1. Open agents.pgbhub.com and sign in with your PGBHub account.
2. Open AI keys. Under Add a key choose the Provider, paste your Key and choose Check and save. Then choose Test connection and look for "Test passed: your agents can reach a model."
3. Open Settings and read the Safety card once. It lists the rules that apply to every agent.
4. Look at the top bar. You will use Agents, Runs, Approvals and Usage throughout the labs. Reports summarises how your agents are doing.

How to read each lab

Every lab has the same shape:

• Goal: what you will have at the end.
• Steps: numbered, matching the real screens.
• Check your work: how to know it worked.
• What you learned: the idea to carry to any platform.

Use a small per-run budget while you learn. In Budget, a Per run (USD) of 0.10 is plenty for the early labs.

Try it yourself

Write a design brief for a "weekly reading digest" agent that you might build one day. Do not build it yet. Keep the brief; lesson 2 shows how to turn a brief into instructions.

Recap

• A design brief of goal, input, tools, output, approvals and limits comes first.
• Your AI key and the Safety card are the prerequisites.
• Each lab: goal, steps, check your work, what you learned.