parallel-execution

Parallel execution lets you run independent tasks concurrently in a single assistant message using the Task tool with run_in_background: true. This reduces total turnaround time for workflows like code review, testing, and documentation by executing them simultaneously instead of sequentially.

Sequential execution completes one task before starting the next, multiplying total time. Parallel execution launches all tasks at once, reducing total time to the slowest task's duration. This skill enforces all Task calls in one message to achieve true parallelism for concurrent workflows.

Yes, this skill orchestrates multiple subagents in parallel across independent analyses like code review, testing, verification, and documentation. Patterns include dynamic subagents, directory-based, and perspective-based parallelization with guidance for prompt design and result synthesis.

After launching parallel tasks, use a TaskOutput mechanism to collect results as they complete. The skill provides patterns for synthesizing results into a cohesive plan, with TodoWrite integration guidelines for status tracking and conflict resolution.

This skill includes error handling patterns for parallel workflows. When a task fails, the framework lets you handle the error without blocking other concurrent tasks, ensuring synchronization and recovery without stopping the entire workflow.