parallel-subagents

Orchestrating parallel subagents in Claude Code requires explicitly using the Agent tool with isolation controls to launch independent investigations. This Skill orchestrates multiple Agents concurrently, enabling safe multi-file analysis by decomposing tasks into independent subtasks with clear isolation guarantees.

Use concurrent agents when subtasks share no state and can be fully decomposed across independent files or components. Sequential execution is preferred when dependencies exist between tasks, ensuring safe task decomposition and deterministic result merging.

Parallel execution merges results deterministically by aggregating findings from multiple subagents, detecting contradictions, and consolidating insights into unified conclusions. This enables concurrent validation and parallel code reviews to produce integrated outputs.

Yes, parallel subagents support concurrent testing and multi-file code validation by launching multiple Agent tool instances simultaneously. Each subagent operates within isolated context boundaries, enabling independent verification across different components without interference.

Parallel agent orchestration is limited by context boundaries and does not perform privileged system changes. It requires explicit Agent tool usage with isolation controls, and tasks with shared state or dependencies cannot be safely parallelized.