Repository navigation
Subagent sessions cannot be prompted. Back to main session. #21458
Description
Question
Why was the Windows version of the sub-agent changed to "Subagent sessions cannot be prompted. Back to main session."?
I believe in some cases, the subagent should be guided.
OpenCode version
1.4.0
Metadata
Metadata
Assignees
Labels
No labels
Activity
github-actions commentedon Apr 8, 2026
This issue might be a duplicate of existing issues. Please check:
If your question is specifically about the Windows behavior or a recent change that introduced this message, please provide more context (e.g., version where it changed, what behavior you expected) so maintainers can investigate.
FQXCS commentedon Apr 8, 2026
I believe in some cases, the subagent should be guided. Just like the previous version
OpenCode version
1.4.0
cpkt9762 commentedon Apr 8, 2026
For anyone hitting this from the error recovery angle (subagent session halted after provider failures / retry exhaustion, composer now locked so you can't type "retry" to resume it) rather than the steering angle this issue discusses — I opened #21525 to track that case separately.
The distinction matters because the "you weren't actually talking to the subagent, it was routed through the primary agent" reply (valid for steering) doesn't apply when the session is already in an error state and there is no live primary-agent turn to route to.
MicroCats37 commentedon Apr 8, 2026
Hello, good morning.
I am currently working on designing a sub-agent-based architecture, where tasks are delegated according to specifications stored in a database. Additionally, every change, thought, or design decision is recorded alongside the executed tasks.
However, I’ve encountered a significant issue: due to model limitations, there are cases where the agents start producing erratic or irrelevant text. This negatively impacts the architecture, especially since it is optimized for token efficiency.
In such situations, the current flow gets interrupted. The main agent has to stop the process and create a new sub-agent with a clean context. Then, this new sub-agent needs to request all the necessary data again in order to resume the task, which introduces overhead and inefficiency.
Because of this, it would be very useful to have an option to enable or control the prompt directly within sub-agents in certain scenarios. This would provide better control over their behavior and help prevent unnecessary disruptions in the workflow.
rekram1-node commentedon Jun 23, 2026
Automated: closing question issues.
GitHub Issues are for bug reports and feature requests only — questions / support are no longer tracked here.
If this is a bug or feature request, please open a new issue with the appropriate template.