subagent-driven-development

A two-stage code review workflow first checks spec compliance and then evaluates code quality after each task. This structured process ensures implementation plans meet requirements and maintain high standards through guided iterations and escalation paths.

This subagent-driven development workflow is suitable for plans with mostly independent tasks that benefit from isolated context. It is designed for rapid, guided iterations where task dispatch and strict spec compliance are required within the current session.

Subagent-driven development coordinates distinct prompts for the implementer, spec reviewer, and code quality reviewer. This separation enforces sequential validation, enables self-review, and provides clear escalation paths if quality thresholds are not met.

You should not use a fresh subagent per task when implementation plans lack mostly independent tasks. This workflow relies on isolated context and rapid iterations, so tightly coupled tasks requiring continuous shared context may not benefit from this dispatch approach.