parallel-dispatch

Parallel task execution across multiple agents is coordinated by orchestrating dispatch with isolated workspaces, ensuring independent components are developed concurrently to maximize throughput and avoid bottlenecks.

Use parallel dispatch when planning a large feature with several independent components that can be worked on concurrently, requiring clear inputs and outputs and isolation to prevent data dependencies between tasks.

Subagent worktrees provide isolated workspaces for each parallel agent, preventing data dependencies by ensuring concurrent task execution does not interfere with shared states before the defined merge and verification strategy.

The best way to merge parallel agent results safely is by applying a defined merge and verification strategy during the execution phase, ensuring deterministic results and safe integration of independently developed components.

Yes, parallel dispatch supports both subagent and non-subagent modes, providing explicit guidance for each path to orchestrate concurrent tasks while maintaining structured task definitions and verification steps.

Parallel task execution requires clear inputs and outputs to ensure deterministic results, allowing isolated agents to process independent components concurrently without creating data dependencies during the merge phase.