Skip to main content
Tasks that exhaust all retry attempts are moved to a dedicated Failed status. This gives you clear visibility into what went wrong and easy recovery options.

​
Visual Indicators

The Failed status is clearly marked with red indicators across every view:
A dedicated red Failed column appears on the Kanban board. Cards show the retry count and failure reason.

​
Task Lifecycle

Backlog ──▶ In Progress ──▶ Testing ──▶ Done
                │               │
                ▼               ▼
             Failed ◀───────────┘
                │
                ▼ (manual action)
             Backlog (retry count resets)
A task reaches Failed status when:
  1. The agent process crashes or exits with an error
  2. The task has already used all configured retry attempts
  3. No more retries are available

​
Recovery

Review the failure

Open the task to see the terminal output and understand what went wrong. Check the last tool calls and error messages.

Fix the issue

Common fixes: update the task description with more context, change the model to Opus for complex tasks, fix a broken dependency, or resolve a git conflict.

Retry the task

Move the task back to Backlog or In Progress. The retry counter automatically resets to 0.
If a task keeps failing, try breaking it into smaller sub-tasks with dependencies. Complex tasks have a higher success rate when decomposed.

​
Conditional Dependencies

Failed tasks can trigger on_failure dependencies. This lets you build recovery workflows:
  • Task A fails → Task B (debug/fix task) starts automatically
  • Notification webhook fires with failure details
  • A simpler version of the task can be attempted with a different model