subagent-driven-development

You isolate task reasoning by dispatching a fresh subagent per task, which prevents context pollution across the session. This approach ensures each task is handled with a clean slate while maintaining automated task tracking throughout the workflow.

A two-stage review workflow evaluates each task through spec compliance and code quality checks performed by dedicated reviewer subagents. This automated process validates that implementations meet requirements before code is merged, iterating until all tasks are complete.

You orchestrate sequential and parallel task workflows by using in-session orchestration to dispatch implementer subagents across independent tasks. This maintains a continuous session while supporting both execution flows for multi-task initiatives.

Yes, you can use defined prompt pipelines that dispatch dedicated spec reviewers and code-quality reviewers for each task. These automated pipelines ensure implementations satisfy compliance and quality standards before finalizing the code.

Use fresh subagents for task execution when your plan involves mostly independent tasks that require isolated reasoning and rapid iteration. Spawning a new subagent per task prevents context pollution and maintains focus for complex, multi-task initiatives.

In-session subagent orchestration is limited to tasks that are mostly independent and require isolated reasoning. It is not suited for highly interdependent tasks where shared context across steps is critical, as each subagent operates with a fresh context.