Things worth trying, grouped by what you are doing.
The tools are self-describing, so you rarely need to name them. Ask for the outcome and let the agent pick. These are grouped by what you are trying to get done.
Build a popup from scratch
“Build an exit-intent email capture popup for example-shop.com”
“I want a spin-to-win wheel for my Black Friday sale on example-shop.com”
“Create a free-shipping bar for example-shop.com and keep it subtle”
The agent asks a couple of questions about the store and the offer first — that is the design thread doing its job, not a stall.
Recreate something you have seen
“Here's a screenshot of a competitor's popup. Rebuild it for example-shop.com”
“Recreate this popup but match my brand colours”
Upload the image and the agent uses upload_reference_image before building.
Change one that exists
“Make the headline shorter and move the button above the image”
“Change the mobile version so the form fits without scrolling”
“Try a version with a 15% discount instead of 10%”
Look at what you have
“Which campaigns do I have on example-shop.com?”
“Show me the HTML of the last popup you built”
“Which OptiMonk account are you working in?”
What it cannot do yet
Taking a campaign live. The agent builds and creates campaign drafts; publishing is done by a person in the app. Reporting numbers are not available to an agent either. Both are waiting on the same piece of work.