subagent-driven-development

Break your implementation plan into isolated tasks and dispatch a fresh subagent for each one. This approach keeps work focused and isolated, ensuring continuous plan execution without pausing for human check-ins.

Subagent-driven development works by running a two-stage review loop after each task: first verifying spec compliance, then checking code quality. Each completed task also requires a self-review and task tracking before moving forward.

The best way to implement unrelated feature changes is dispatching independent subagents per task. This allows you to implement, review, and finish each isolated step cleanly before moving to the next one without losing context.

Yes, a spec compliance check is required before the code quality review. This two-stage review loop ensures each task meets the original implementation plan specifications before assessing the overall code quality.

Yes, you can use subagents for continuous plan execution without human check-ins. This Skill is designed for multi-step development work where tasks can be completed in sequence autonomously.

You should not use isolated subagents when your development plan involves tightly coupled steps that cannot be divided into isolated tasks. This approach requires work that can be separated into independent implementation tasks, questions, and sequential reviews.