subagent-driven-development

Subagent-driven development splits implementation plans into isolated tasks, dispatching a fresh subagent for each task to keep context clean and maintain focus throughout sequential execution.

Two-stage review enforces spec compliance first, then code quality, after each subagent implementation task. This sequential oversight ensures the controller validates deliverables without guessing or losing track of progress.

The workflow handles question-asking, rework, and blocker escalation so the controller can keep moving. It routes questions back to the controller instead of having subagents guess, ensuring spec compliance is maintained.

Yes, it applies to same-session execution of implementation plans, especially multi-step feature work, bug fixes, and other changes that benefit from isolated context and sequential oversight across independent tasks.

Yes, this workflow is most useful when you already have a clear plan and need to deliver multiple related code changes in the same session, assigning each task to a fresh subagent with enforced review.

The workflow requires todo tracking alongside per-task implementer dispatch and two-stage review, ensuring the controller maintains visibility over task completion and escalation handling throughout the implementation plan.