subagent-driven-development

You execute implementation plans by dispatching fresh subagents per task. This ensures isolated context and fast iteration, with a structured plan-to-task workflow that automates orchestration and integration checks.

The two-stage review process first checks spec compliance to verify task alignment, then evaluates code quality. This enforces consistent quality before task integration is validated.

You parallelize task dispatch by assigning independent subagents per task. This enables parallel work while maintaining clean context, followed by automated two-stage reviews for final integration validation.

Yes, the workflow supports iterative feedback loops during plan execution. Subagents receive feedback on spec compliance and code quality, iterating until tasks are ready for integration validation.

Yes, you must provide an implementation plan document as input. The system reads this plan and executes it by dispatching per-task subagents to maintain isolated context.

Using fresh subagents for per-task delegation requires a well-structured plan to function correctly. Context is isolated per task, so dependencies between tasks must be managed through the final integration validation phase.