subagent-driven-development

You can execute multi-file feature delivery plans without context overload by delegating independent implementation tasks to fresh subagents, ensuring each work item stays isolated. This requires precise context packaging for every delegated task to maintain focus.

A two-stage review process for task delegation verifies specification compliance first, then checks code quality before marking a task complete. This enforces a consistent review loop for every independent work item delivered by subagents.

Yes, subagent task isolation applies to iterative refactors and staged development branches by giving each isolated work item a clean context. This makes coordinated multi-file changes easier to debug and track through explicit completion monitoring.

Coordinate code review and spec review for independent implementation tasks by running self-review and specification compliance checks before code-quality review. Explicit completion tracking ensures every delegated task passes both stages before completion.

You should not use subagent-driven development for plan execution when tasks are highly interdependent rather than isolated, or when a plan spans multiple sessions. It is designed for single-session execution plans with independent work items.