subagent-driven-development

Automate development plan execution by dispatching independent subagents for each task, followed by a two-stage review process checking spec compliance and code quality. This manages task flow and iteratively fixes code based on review feedback.

Subagent-driven development executes implementation plans by delegating individual tasks to fresh, isolated subagents. It ensures structured development by using specific prompt templates for implementer, spec reviewer, and code quality reviewer roles.

Yes, subagent-driven development provides context isolation by assigning each task to a fresh subagent. Subagents receive precisely the context they need, preserving the main agent's context during complex feature implementation.

Ensure spec compliance using a two-stage review process with separate checks for specification adherence and code quality. After a subagent implements a task, a spec reviewer subagent verifies compliance before proceeding to the next task.

Start automated plan execution by providing a detailed implementation plan document, such as a markdown file at docs/plans/new-feature.md. The system systematically executes each step from initial implementation to final code review autonomously.

Avoid autonomous subagents for task delegation when your implementation plan lacks detailed steps or when tasks require continuous shared context rather than isolated execution and rigid two-stage spec and code quality reviews.