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
Bug: Empty task output breaks fallback system #29054
When a task completes with no text output (e.g. rate-limit or empty response), the tool currently returns an empty string instead of failing. This breaks the fallback system and prevents retrying with the next configured model.
Note: I've opened a related issue (#29470) tracking a sibling edge case: when the LLM provider API connection hangs indefinitely without returning any data or throwing an HTTP error, it entirely bypasses the fallback chain since the stream never completes.
Activity
github-actions commentedon May 24, 2026
This issue might be a duplicate of an existing issue. Please check:
TaskToolsilently returns an empty string when the subagent result has no text part (the?? ""fallback), and discusses the same problem of empty<task_result>breaking downstream behavior.niStee commentedon May 27, 2026
Note: I've opened a related issue (#29470) tracking a sibling edge case: when the LLM provider API connection hangs indefinitely without returning any data or throwing an HTTP error, it entirely bypasses the fallback chain since the stream never completes.
niStee commentedon May 27, 2026
Closing as a duplicate of the earlier issue #24447! PR #29048 implements the fix for this.