subagent-driven-development

Iterative development with AI agents is automated by dispatching independent subagents per task with a two-stage review process. This Skill breaks down your plan into manageable tasks, assigning a fresh subagent to each for focused implementation and structured feedback.

The two-stage code review process enforces a spec compliance review followed by a code quality review for every task. It ensures each piece of your implementation plan meets standards through managed feedback loops between implementer and reviewer subagents.

You execute an implementation plan by pointing the Skill to your plan file path. It automatically decomposes the plan into independent tasks, dispatches a fresh subagent for each, and manages the iterative review cycles within a single session.

Yes, you can use this subagent dispatch approach for feature development involving API endpoints, UI components, or database schema changes. Each part of the feature is handled as a separate task, ensuring every piece is reviewed and approved before moving to the next.

Using independent subagents for task management ensures focused implementation per task and prevents context degradation. Fresh subagents maintain quality through structured spec compliance and code quality reviews, facilitating rapid iteration on your development plan.

A limitation of subagent-driven iterative AI development is that it requires a detailed implementation plan file as input. Furthermore, the two-stage review process operates within a single session, meaning task execution and feedback loops depend heavily on the initial plan quality.