Isolate the failing component
Ask the agent to narrow down exactly which piece of the app is broken, by checking the pieces separately, before changing anything.
Narrow a failure to one module, service, or step by testing parts independently: calling an API directly, feeding in known-good inputs, or disabling pieces one at a time until the fault is contained to a single component.
When a string of holiday lights goes dark, you test the bulbs one at a time until you find the dead one, instead of buying all new lights.
When a feature spans the screen, the server, the database, and outside services, "it's broken" could mean any of them. Isolating the failure keeps your agent from changing four places to fix one.
I isolated the failure: the API returns correct data, and the chart draws sample data fine. The bug is in the date formatting between them.
The sales chart is empty. Isolate the failing component first: check the API response, the data conversion, and the chart separately, and tell me which one is broken.
Changing several things at once to see if the problem goes away. Even if it does, you won't know which change fixed it.