Good comment from a user: we should make it more clear what the success/failure cases of automations are so that when it fails it actually gets registered as a failure

i think there are two types of failures 1. systems around the agent that fail 2. agent failing at the task both have better visibility in canvas as of recently; task outcomes help with the second point now, the user can define what success like any other gaps we can think of here?

I think the comment was about the gap between “the user can define what success looks like” and “the user knows to define what success looks like”. So if we have a UI or a conversation-based interface to create automations, then success criteria should be part of that