orchestrate-parallel-work

Splitting coding tasks into parallel sub-agents safely requires decomposing the work into non-overlapping scopes with explicit ownership. The orchestrator defines per-worker plans, optional worktree isolation, and verification commands to maintain disjoint boundaries.

Task decomposition for parallel work orchestration is the process of dividing a request into isolated slices. Each slice is assigned to a worker with explicit ownership and centralized integration managed by an orchestrator.

To prevent file conflicts when running parallel sub-agents, you enforce non-overlapping scopes and use optional worktree isolation. The orchestrator assigns explicit ownership to each worker, ensuring they only modify designated slices.

Yes, you can use worktree isolation for parallel coding tasks. The orchestrator supports optional worktree isolation to keep worker environments separate, preventing overlap while each sub-agent executes its specific plan.

You should use an orchestrator for delegated sub-agent work when the user explicitly asks for delegation and the task can be split into non-overlapping scopes. The orchestrator provides centralized integration and verification for the parallel slices.

You verify the results of parallel sub-agents through orchestrator-led integration. The orchestrator defines specific verification commands during task decomposition to validate the centralized integration of disjoint worker slices.