hey has anybody tried to contribute to the new terminal bench 3 open call and tested with openhands agent? any tiny bit of insight would be super helpful

I just added the OpenHands SDK adaptor to Harbor (the library behind terminalbench) yesterday, and was going to test it out today!

wait you did? that is great! what does that change in terms of implementation for folks running trials with harbor?

It should be trivial now?

ohhhh right! i remember now, the previous agent in installed_agents was not an openhands sdk v1 agent. i can try running it on this computational bio hellscape of a task that i am working on, if you want

that’d be great!