delegation

To coordinate multi-agent tasks across worktrees, you dispatch implementation tasks to subagents or agent teams with strict per-task worktree isolation and TDD-driven validation. The system orchestrates preparation, dispatch, and monitoring phases while recording provenance for traceability.

Yes, you can parallelize a feature across multiple subagents by dispatching tasks into isolated worktrees with separate feature branches. The orchestration mode enforces deterministic prompts and quality gates, ensuring each parallel task runs independently with CI checks.

TDD-driven validation in agent teams orchestration works by enforcing quality gates during the monitoring phase. It ensures deterministic execution across dispatched tasks, validating that implementation meets test requirements before progress gating allows completion or triggers recovery paths.

The recovery path for a failed subagent task uses fix-mode and operator gates. When validation fails during monitoring, the orchestration enforces deterministic execution and provides these recovery paths to handle errors without losing per-task context isolation or provenance traceability.

Yes, separate worktrees are required for each task to maintain strict context isolation. The orchestration enforces per-task worktrees and deterministic prompts, ensuring parallel implementation across subagents or agent teams does not interfere and provenance remains traceable.

The best way to monitor progress across dispatched agent teams is through the orchestration's monitoring phase. It handles progress gating and records provenance, providing operator gates to oversee deterministic execution and ensure TDD-driven validation passes before task completion.