Copy This Prompt - Mapping Your Agent's Loop

Prompt

Help me think through how an agent would handle a task in my product, step by step.

THE TASK: [what you want the agent to accomplish]
SYSTEMS IT COULD REACH: [list what data and actions exist]
WHAT A HUMAN DOES TODAY: [walk through it, roughly]

Please:
1. Write out a plausible step-by-step trace for a typical case, what the agent thinks, what tool it calls, what it observes, and why it continues
2. Now write a trace for an awkward case, where the first thing it tries doesn't work
3. How many steps does each take? Where does the number vary most?
4. Name three points in these traces where the agent could plausibly choose wrongly, and say what it would do instead
5. What should the give-up condition be, and what should the agent hand back to the human when it triggers?