Does anyone else have their agent just... keep going forever?
I've been using OpenCode recently and I've noticed something I don't really understand yet.
Sometimes I give it a task and it just keeps working. It'll read files, make changes, run something, read more files, go back and check something again... and it feels like it's stuck in a loop without actually getting closer to finishing.
Usually I end up hitting Esc, then telling it to continue, and somehow it picks things up and finishes the task.
I'm still pretty new to agentic coding, so I'm wondering if I'm doing something wrong here.
Is this usually caused by the model, the agent's tool loop, the way the task is written, or something else?
And if you've run into this with OpenCode (or other coding agents), how do you usually deal with it without having to babysit the agent every few minutes?