subagent-driven-development

To automate task dispatch to sub-agents, you provide a plan text to the main agent, which systematically distributes each self-contained task to a fresh sub-agent for isolated execution within a single session. This approach enables rapid iteration and prevents context pollution.

The two-stage review process for sub-agent code generation first evaluates specification compliance to ensure the implementation meets requirements, and then, if passed, conducts a code quality review to enforce end-to-end quality standards after each task is completed.

You prevent context contamination when delegating multiple development tasks by assigning each task to a dedicated, fresh sub-agent. This task-level dispatch ensures sub-agents remain independent, avoiding conflicts and maintaining high focus and traceability.

Yes, you can execute iterative development plans with isolated sub-agents in one session by applying this workflow. It supports rapid iteration by dispatching tasks to independent sub-agents, followed by two-stage reviews and branch finalization steps.

You should use a sub-agent driven workflow for task automation when your tasks are self-contained and require in-session execution. It is ideal for plans needing rapid iteration, strict specification compliance, and code quality enforcement without context interference.

To finalize tasks after sub-agent execution and review, the main agent enforces a finishing step using dedicated sub-agents. This finalization ensures end-to-end quality is maintained after all two-stage reviews and iterative executions are complete.