dispatching-parallel-agentslisted

# Dispatching Parallel Agents ## Overview You delegate tasks to specialized agents with isolated context. By precisely crafting their instructions and context, you ensure they stay focused and succeed at their task. They should never inherit your session's context or history — you construct exactly what they need. This also preserves your own context for coordination work. When you have multiple unrelated failures (different test files, different subsystems, different bugs), investigating them sequentially wastes time. Each investigation is independent and can happen in parallel. **Core principle:** Dispatch one agent per independent problem domain. Let them work concurrently. ## The boundary: not the tasks of a plan **This skill does not reach the tasks of a plan under execution.** There, superpowersplus:subagent-driven-development governs, and its rule is unconditional: never dispatch multiple implementation subagents in parallel. That rule wins inside a plan run, and nothing here softens it. The boundary is easy to miss because this skill fires on its own description — no skill routes to it — so it is reachable from inside a plan run, where it does not belong. Two reasons it does not: - **Edit conflict.** Plan tasks are decomposed to be reviewable one at a time, not to be disjoint on disk. Two implementers in the same file is a merge nobody asked for, discovered after both finished. - **The next task's review depends on the last one finishing.** A task reviewer