review-codebase

To audit a codebase for architectural drift, dispatch parallel specialized reviewers to assess substrate alignment, code structure, library-native patterns, and agent-readiness, then synthesize their findings into a thesis-led report with prioritized remediation steps.

The best way to check subsystem cohesion is auditing named subsystems against normative documentation to verify enforcement of invariants and correct placement of architectural seams, ensuring the system scales without relying on founder memory.

You cannot run a valid codebase review with sparse or inconsistent documentation, as the parallel review process halts early if underlying architectural specs and invariants are too insufficient to support a valid assessment.

Parallelized codebase review dispatches four focused reviewer agents simultaneously to evaluate substrate alignment, code structure, library-native patterns, and agent-readiness, synthesizing their findings into a single thesis-led report with actionable recommendations.

You need an architecture review when planning a major refactor, onboarding to a legacy codebase, or auditing for hidden architectural drift before high-stakes changes to ensure the system can scale development without relying on founder memory.

Codebase review works for pre-change health checks by applying named subsystem cohesion audits to validate architectural consistency before major feature work, ensuring intended system boundaries and invariants are correctly enforced.