subagent-driven-development

To prevent context leakage when executing large implementation plans, dispatch a fresh subagent for each task. This isolates execution contexts and avoids cross-task contamination, ensuring consistent quality and preventing context window overflow.

Automating spec compliance and code quality reviews is best achieved through a two-stage review process. This mechanism enforces spec alignment first, followed by a separate code quality review, catching issues early across development workflows.

Orchestrate implementation plans for parallel execution by dispatching a fresh subagent per task. This per-task delegation maximizes parallelism and provides isolated execution contexts for each independent task within the workflow.

Yes, subagent delegation requires a pre-defined implementation plan. The system integrates with plans created by the writing-plans skill, parsing the plan to dispatch a fresh subagent for each task and deliver repeatable results.

Subagent-driven development applies a two-stage review process to enforce disciplined quality control. A spec reviewer ensures alignment with requirements, and a separate quality reviewer checks maintainability, catching issues early before integration.

Subagents are suitable for independent tasks where isolated execution contexts are beneficial. For dependent tasks, cross-task contamination prevention remains, but parallelism is limited by the sequential nature of the dependencies.