@calm-artist-7206 @limited-student-10747 Here is a new proposal for fixing subagents in the remote workspace. It is a minimal change and includes only one breaking API change. Note: One test is failing, but it is unrelated to the changes in this PR; it appears to be a CI-specific issue. https://github.com/OpenHands/software-agent-sdk/pull/2323

Thank you! TBH I skimmed a bit, but conceptually it's fine imho, we could do it like this and we'll see. I set off the agent to try examples in advance No idea what test was it, but we can rerun that workflow if it's flaky

It seems to be the case. They are passing locally