subagent-driven-development

To delegate multi-task plans without context contamination, dispatch a fresh subagent for each task. This isolated context approach ensures clean execution and prevents cross-task interference during your implementation workflow.

A two-stage subagent review process evaluates completed tasks sequentially: first checking spec compliance to verify requirements, then assessing code quality. This enforces quality and traceability across multi-task projects before marking tasks complete.

You need isolated subagent delegation when coordinating multi-task implementation plans where tasks are largely independent. It provides clean, isolated context for each subagent to prevent cross-task contamination and ensure traceability.

To execute planning-driven delegation, dispatch a new delegate_task for each task in your plan. After implementation, run the two-stage review process to check spec compliance and code quality before marking tasks complete.

Subagent-driven development is designed for multi-task plans where tasks are largely independent. Tightly coupled tasks requiring shared context may not benefit from the isolated subagent approach and clean context separation.

Inconsistent code quality often occurs when skipping the two-stage review process. Enforcing both spec-compliance and code-quality reviews for each delegated subagent task ensures consistent quality and traceability across the project.