dispatching-parallel-agentslisted

# Dispatching Parallel Agents Use this skill when multiple independent problems can be split into separate agent tasks and run concurrently. ## Why This Skill Exists This skill is about parallelism, not plan execution. Use it when one agent would otherwise investigate or implement several unrelated things in sequence even though they do not depend on each other. ## When to Use This Skill Use this skill when: - two or more failures have different root-cause areas - multiple subsystems need investigation and they do not overlap - several tasks can be completed without editing the same files or relying on the same intermediate state - parallel execution will save time without increasing coordination risk Examples: - three failing test files in unrelated subsystems - separate debugging tasks in different packages - independent code review or research tasks ## Do Not Use This Skill When - tasks depend on each other - fixing one issue may change the others - agents would touch the same files or shared mutable state - the work needs one coherent view of the entire system first - a plan requires strict sequential task order In those cases, use `executing-plans` or `subagent-driven-development` instead. ## Core Pattern ### 1. Identify Independent Domains Before dispatching anything, group work by problem domain. Each parallel task should have: - a narrow scope - a clear boundary - minimal overlap with the others If you cannot explain why two tasks are independent, do