Repository navigation
Limit Subagent Tree Depth #2906
Description
I'd like to be able to limit subagents from creating further subagents too deeply. Often I'll find them creating a deeply nested tree which usually means they're duplicating work that was done higher in the tree, or at the very least should be orchestrated by the parent agent explicitly instead of the subagent deciding to "analyze this github repo" on its own which in turn creates its own set of recursive trees.
Perhaps a default limit of 2-3 would make sense. In my experience, once it goes any further than that, it's usually an indication of a problem.
It'd be easier to catch these kinds of issues if sub-sub-agents were visualized. Currently only the initial call to the subagent is shown; ideally nested tool calls under that agent would be visually shown and indented to make it clear what's actively being done without having to switch to that agent's session.
Activity
github-actions commentedon Oct 1, 2025
This issue might be a duplicate of existing issues. Please check:
Feel free to ignore if none of these address your specific case.
rekram1-node commentedon Oct 2, 2025
yeah that makes sense
anis-dr commentedon Mar 9, 2026
This is actually a problem, GPT 5.4 especially loves to delegate work to subagents and it never stops.
github-actions commentedon May 9, 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.