parallel-task

To execute dependent tasks in parallel, you provide a plan file to the orchestration skill. It derives tasks with depends_on relationships and launches unblocked subagents in waves to maximize speed while preserving correctness.

Wave-based orchestration is a coordination method that groups unblocked tasks to run in parallel. It ensures tasks with depends_on relationships wait for prior results, making it ideal for multi-step pipelines like plan execution and data processing.

You can run a specific subset of tasks by targeting the planning document. The orchestration skill supports running targeted work by executing only the specified subset while still honoring depends_on relationships and updating logs.

Yes, you can use this skill to log progress after each task completes. It automatically updates the plan docs with per-task logs after execution, ensuring reliable progress tracking across all parallel subagents.

You need parallel task orchestration when executing multi-step pipelines where tasks depend on prior results. If your planning, testing, or data processing workflows require coordinating dependent tasks across multiple subagents, this approach prevents errors and increases speed.

A limitation of coordinating subagents in waves is that tasks must have clearly defined depends_on relationships in a plan file. Without structured task graphs, the orchestration cannot determine which tasks to unblock and run in parallel.