subagent-driven-development

Two-stage review prevents spec drift by running a spec compliance check first, then a code quality review, for each task executed by an isolated subagent. This structured process iterates fix-and-recheck loops until the implementation is approved before integration.

Subagent-driven development is a workflow technique that bounds context by spawning a fresh subagent per task during parallel implementation. It reduces cross-task confusion by isolating changes and verifying each task independently before performing a final integration review.

Verify code quality and spec compliance by applying a two-stage review process per task: run spec compliance checks before code quality reviews. This sequence ensures the implementation scope is correct first, then validates code solidity, iterating fix-and-recheck loops until both pass.

Subagent-driven development works for large refactors and feature development by parsing an implementation plan once and dispatching isolated subagents for each extracted task. It bounds context to manage complexity, finishing with an integration readiness check and full test run.

Avoid using isolated subagents for implementation planning when changes are highly dependent on each other, as the workflow relies on dispatching largely independent tasks. The overhead of spawning subagents and running two-stage reviews is best suited for multi-task feature development and complex refactors.