subagent-driven-development

It works by dispatching a fresh subagent for each task in your plan, followed by a two-stage review process. The first stage checks specification compliance, and the second evaluates code quality to ensure high standards.

You orchestrate multi-task plans by dispatching a fresh implementer subagent for each task. This isolates the context per task, enabling quick iteration and preventing interference across the broader workflow.

The two-stage review process executes after every dispatched task. It first verifies specification compliance to ensure requirements are met, then assesses code quality to maintain high standards across the independent subagent outputs.

Yes, subagent task dispatch supports parallel execution paths within a single session. It uses plan-driven task dispatch followed by final consolidation to manage multi-task plans effectively.

Use a fresh subagent per task when working with multi-task plans where quick iteration and context isolation matter. This standardized workflow prevents context pollution and maintains spec compliance across independent tasks.