parallel-execution

You can run multiple subagents concurrently by issuing all Task calls in a single assistant message with run_in_background set to true. This parallel execution pattern allows independent tasks to execute simultaneously, reducing overall execution time.

Parallel task orchestration coordinates multiple subagents using the Task tool to run independent sub-tasks concurrently. It applies to data processing pipelines and automation workflows where sub-tasks can execute simultaneously instead of sequentially.

You execute non-blocking background tasks by launching each Task with run_in_background: true in a single assistant message. After launching them in parallel, you collect the TaskOutput IDs and synthesize the results once the independent tasks complete.

Yes, parallel subagent execution supports Task-Based, Directory-Based, and Perspective-Based Parallelization patterns. These patterns allow you to coordinate independent data processing tasks across different codebases, directories, or analytical perspectives concurrently.

You should avoid concurrent subagents when tasks are dependent on each other, as parallel execution requires independent sub-tasks. This skill is specifically designed for automation workflows where sub-tasks can execute without waiting on each other's outputs.