Is it normal that sometimes we need to pause and restart the conversation in order for the agent to process our latest message immediately? Otherwise, the agent continues with its unnecessary internal reasoning and struggles because it doesn’t take your message into account, even though your message clarifies everything. Is it simply finishing the last inference request, or is some kind of loop happening? Other agents interrupt the current inference request or provide a

/steer

command.

We had a long thread about interupting and steering here: https://openhands-ai.slack.com/archives/C06QT0AGY4W/p1788354734889099

Thanks @average-processor-75195, I read the thread. So /steer is what I want, and that's what OH is dellivering

I think so. One other issue I see is with subagents that can feel slow, because they don't give any progress.

agreee

Also the system prompt encourages the agent to be very complete. I want to have an agent profile that does the basics and moves along quicker for some tasks, like doing some basic research tasks.

I know it is possible to define a custom agent with a custom system prompt and tools, but I'm not sure if there is a way to tell each sub-agent which agent to use

I think we are adding more tools for agent profiles. You also can just spin up a second agent instead of a subagent.

What I do is spawn a normal agent and give them an initial message with a .md to read

Ahh, that makes sense, I just wanted a custom agent, so I wouldn't always ahve to the initial for some use cases, but totally following you.