implement-plan

To orchestrate plan execution with parallel subagents, you provide a plan file or plan text. The system analyzes tasks, groups them into parallel batches, and delegates execution to plan-implementer subagents while preserving overall context.

Task decomposition and batching work by analyzing dependencies and grouping tasks into parallel batches. The system ensures tasks touching the same files are not run in parallel and enforces a maximum of four concurrent subagents during execution.

You do not need a specific plan file format to automate task implementation. You can provide either a structured plan file or plain plan text, and the orchestrator will identify discrete tasks, dependencies, and ambiguities for implementation.

When a subagent fails during parallel plan execution, the orchestrator handles the failure with retries or reassignments. It delegates the task to another plan-implementer subagent and performs final verification to ensure successful implementation.

The maximum number of concurrent subagents for task orchestration is four. This concurrency limit ensures stable parallel execution while preventing tasks that touch the same files from running simultaneously and causing conflicts.