You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Build got stuck for 6h after delegating to explore subagent #23296
I observe something similar with opencode 1.14.x - 1.17.x (current) -- after explore agent finishes, the build agent keeps showing a spinner and never finishes (for 1-5 hours or longer -- until killed by a custom watchdog script).
Since the LLM is served locally by llama.cpp, I can see that there's no LLM activity, so it's a hang.
Setup : a Debian Linux docker container running opencode, working with a local llama.cpp server serving a qwen3.6 instruct model.
Task: opencode is invoked by by a script which traverses a tree of library modules, one call per each module; there is a "master" opencode session listening at a port, and the run session attach to it to do work; the idea is that it is possible to do an opencode attach -c to the same server and see what the current run process is doing.
Sadly sometimes the run session hangs, and it is most annoying when it can not basically get past a specific module, so restarting the task does not help as it keeps getting stuck at the same spot.
I was trying various workarounds, e.g. starting a new opencode process for each task, but it did not help much.
So yes, seems like there's a bug and yes, a sub-agent timeout setting (15080) would likely help a lot.
Activity
github-actions commentedon Apr 18, 2026
This issue might be a duplicate of existing issues. Please check:
If your issue is distinct from all of the above, please add details explaining how it differs.
gf-mse commentedon Jun 20, 2026
I observe something similar with opencode 1.14.x - 1.17.x (current) -- after explore agent finishes, the build agent keeps showing a spinner and never finishes (for 1-5 hours or longer -- until killed by a custom watchdog script).
Since the LLM is served locally by
llama.cpp, I can see that there's no LLM activity, so it's a hang.Setup : a Debian Linux docker container running opencode, working with a local llama.cpp server serving a qwen3.6 instruct model.
Task: opencode is invoked by by a script which traverses a tree of library modules, one call per each module; there is a "master" opencode session listening at a port, and the run session attach to it to do work; the idea is that it is possible to do an
opencode attach -cto the same server and see what the current run process is doing.Sadly sometimes the run session hangs, and it is most annoying when it can not basically get past a specific module, so restarting the task does not help as it keeps getting stuck at the same spot.
I was trying various workarounds, e.g. starting a new opencode process for each task, but it did not help much.
So yes, seems like there's a bug and yes, a sub-agent timeout setting (15080) would likely help a lot.
github-actions commentedon Aug 20, 2026
To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.