subagent-driven-development

Subagent-driven development orchestrates complex plan execution by dispatching fresh subagents to implement individual tasks and perform per-task reviews. It isolates context to reduce pollution and accelerate iteration while maintaining spec alignment.

A two-stage review enforces spec compliance first, then evaluates code quality. This loop runs after each task implementation by a subagent, ensuring the output meets requirements before moving to the next task.

Yes, subagent-driven development supports both sequential and parallel task execution within a single session. It automatically updates TodoWrite to track progress across independent tasks.

Provide the plan text to the controller. It reads the plan file, extracts all tasks with full text and context, dispatches an implementer subagent for each task, and runs a final integration review.

Using a fresh subagent per task isolates context and improves focus. This reduces context pollution from previous tasks, enabling faster iteration and better code quality during complex plan execution.

This approach is designed for plans with multiple independent tasks. If your plan involves tightly coupled tasks requiring shared context or a single simple task, subagent orchestration may add unnecessary overhead.