For instance, I feel like we might already have a status enum that looks a bit like this:

TaskOutcomeStatus = Literal[
    "success",
    "partial_success",
    "blocked",
    "failed",
    "unknown",
]

github.com/OpenHands/software-agent-sdk/pull/4479#… ^ breakdown of the existing pydantic models we already have, seems like none of them address task success Although I agree I'd like to simplify as much as possible, just agent summary text + success/blocked/failed enum should be sufficient structured response is meant to be flexible (and in the future we'd define it on a per automation basis); so the pydantic model introduced here is not baked into the core sdk its a preset we provide as a fallback, so i'm not super worried about introducing it

but yeah, sgtm

But I agree, I think we need to try stuff🤔