subagent-driven-development

Plan-driven development automation is achieved by dispatching a fresh subagent per task, ensuring isolated execution. This prevents context leakage and drift while maintaining strict scope compliance through automated task delegation.

Context drift happens because shared execution sessions accumulate irrelevant information across tasks. Dispatching fresh subagents per task prevents context leakage by keeping each implementation isolated and focused on its specific plan requirements.

The best way to enforce spec compliance and code quality is using a two-stage review process. After dispatching implementers, the workflow verifies specifications first, then performs a dedicated code-quality review before marking tasks complete.

Subagent workflows are designed for plans with multiple independent tasks. Rapid iteration and rigorous reviews improve spec compliance when tasks can be executed in isolated contexts without interdependencies affecting the delegated subagents.

To start orchestrating per-task subagent workflows, provide an implementation plan and activate the first per-task workflow. This begins automatic delegation, parsing the plan into tasks for isolated subagent execution and subsequent reviews.

You should not use subagent-driven development when your implementation plan contains highly interdependent tasks. The workflow relies on dispatching fresh subagents per task for isolated execution, which requires tasks to be independent for rapid iteration.