phased-implementation-review-loop

Multi-step code refactoring requires a disciplined cycle of planning, step-by-step implementation, and independent review. You break the change into testable steps, implement one at a time, and use fresh-context sub-agents to review each step before advancing.

A fresh-context sub-agent review is an independent verification step where a separate agent evaluates your code changes without prior context bias. It identifies High and Medium severity findings that must be fixed before advancing to the next implementation step.

Plan multi-module feature rollouts by breaking them into small, testable steps with clear goals and likely files. Execute each step individually, run focused local tests, and require independent review approval before moving to the next phase.

You need a phased implementation review loop when bug fixes or refactors involve multiple steps that must be verified sequentially. It enforces planning, controlled execution, and independent review to prevent drift and ensure correctness.

Yes, this approach is designed for multi-module refactors and feature rollouts at scale. It provides guardrails to keep large, complex changes sequenced correctly, verified step by step, and reviewed independently before advancing.

When a sub-agent review identifies High or Medium findings, you must fix those issues and repeat the review process. The implementation step cannot advance until the review loop confirms the code is clean.