Problem (one or two sentences)
unable to create/delegate subtasks (multi level) after interruption
Context (who is affected and when)
Everyone and anytime.
Reproduction steps
- Start a task and have it create a subtask via
new_task (level 1).
- From within that subtask, create another nested subtask (level 2) —
confirm multi-level delegation works normally at this point.
- While a subtask is active/delegated, interrupt the session
(e.g. reload VS Code window, restart the extension, or cancel mid-run).
- Resume the task/session.
- Ask the (resumed) parent task to create a new subtask via
new_task again.
- Observe: the tool call fires with correct task details/parameters,
but no actual subtask/delegation happens — instead the UI gets stuck
showing a "Continue" button.
- Click "Continue" — it does not proceed, loops back to the same stuck state.
- Try creating/delegating a subtask again — same stuck "Continue" behavior repeats.
Expected result
After resuming an interrupted task, calling new_task should delegate normally and spawn/execute the new subtask, exactly as it does on a fresh (non-interrupted) session — including multi-level nesting.
Actual result
Actual Behavior: After interrupting and resuming a task that had an active/delegated subtask, the parent task appears to retain stale "delegated" status metadata pointing at a subtask that no longer exists. Any subsequent new_task call is correctly invoked with the right parameters, but delegation silently no-ops and the UI instead shows a "Continue" button. Clicking Continue does not resolve it and re-enters the same stuck state on every following delegation attempt — effectively an infinite loop that permanently blocks further subtask creation for that task until the whole session/task is abandoned and restarted from scratch. Suspected root cause: The "single-open subtask" guard that prevents a parent from spawning two children at once checks a persisted status flag (something like status: "delegated" / awaitingChildId) rather than verifying the referenced child task is actually still alive. When a task is interrupted mid-delegation, the child is torn down but the parent's persisted metadata is never reset back to "active" / cleared. On resume, the parent reloads that stale metadata, the guard thinks a child is still running, and blocks/loops instead of delegating.
Variations tried (optional)
No response
App Version
v3.74.0 (Release)
API Provider (optional)
None
Model Used (optional)
No response
Zoo Code Task Links (optional)
Video of the problem: Link
Relevant logs or errors (optional)
Problem (one or two sentences)
unable to create/delegate subtasks (multi level) after interruption
Context (who is affected and when)
Everyone and anytime.
Reproduction steps
new_task(level 1).confirm multi-level delegation works normally at this point.
(e.g. reload VS Code window, restart the extension, or cancel mid-run).
new_taskagain.but no actual subtask/delegation happens — instead the UI gets stuck
showing a "Continue" button.
Expected result
After resuming an interrupted task, calling
new_taskshould delegate normally and spawn/execute the new subtask, exactly as it does on a fresh (non-interrupted) session — including multi-level nesting.Actual result
Actual Behavior: After interrupting and resuming a task that had an active/delegated subtask, the parent task appears to retain stale "delegated" status metadata pointing at a subtask that no longer exists. Any subsequent
new_taskcall is correctly invoked with the right parameters, but delegation silently no-ops and the UI instead shows a "Continue" button. Clicking Continue does not resolve it and re-enters the same stuck state on every following delegation attempt — effectively an infinite loop that permanently blocks further subtask creation for that task until the whole session/task is abandoned and restarted from scratch. Suspected root cause: The "single-open subtask" guard that prevents a parent from spawning two children at once checks a persisted status flag (something likestatus: "delegated"/awaitingChildId) rather than verifying the referenced child task is actually still alive. When a task is interrupted mid-delegation, the child is torn down but the parent's persisted metadata is never reset back to "active" / cleared. On resume, the parent reloads that stale metadata, the guard thinks a child is still running, and blocks/loops instead of delegating.Variations tried (optional)
No response
App Version
v3.74.0 (Release)
API Provider (optional)
None
Model Used (optional)
No response
Zoo Code Task Links (optional)
Video of the problem: Link
Relevant logs or errors (optional)