Describe what to test; the agent drafts journeys with an objective, an expectation, steps and {{inputs}}.
Journeys are persona-agnostic in wording; personas bind per journey at save time with their own field values. 0 = the generic lane.
The agent emits what happened; the platform derives passed / failed / indecisive. There is no wire status to fake.
Resolve an indecisive outcome once and the ruling is learned per application, so the next run doesn't ask again.
3 flows · 5 runs