team-lead

To coordinate parallel execution of sub-agents during the SDD apply phase, this orchestrator launches multiple @implement tasks simultaneously, manages file reservations, and broadcasts completion to dependent team leads.

The best way to handle file conflicts when dispatching parallel task groups is using file reservation and conflict management, which detects conflicts and serializes tasks within a group when necessary.

You can run dependent task groups in independent mode to execute immediately, or use dependent mode to wait for upstream groups, ensuring reliable completion of the apply phase orchestration.

State management and recovery for task board progress relies on tb_status and SQLite as the single source of truth, persisting incremental progress to Cortex and producing consolidated apply reports.

Parallel dispatched tasks fail during pre-flight checks due to file reservation conflicts or missing dependencies, requiring the retry handling mechanism to manage state recovery and ensure reliable completion.

Yes, this orchestrator supports broadcasting completion to dependent team leads, managing artifact retrieval and releasing file reservations promptly after parallel task groups finish.