Trimming overgrown codebase

This audit command, "audit boundaries," helps trim overgrown codebases resulting from rapid feature development or LLM-generated code. It analyzes layer handoffs to ensure proper separation of concerns, preventing leakage of responsibilities between layers (presentation, domain, infrastructure). The audit identifies presentation data violations, domain infrastructure violations, fat components (overly large components), unauthorized database access (outside the designated data layer), and scattered invariant logic.

Specific findings from the audit include components performing data mutations and cache management in the presentation layer, which should be extracted into separate hooks. It flags components with complex prefetch logic needing extraction. It also highlights business logic being implemented in inappropriate locations (e.g., within handlers). The audit surfaces instances where state mutation and UI logic are intertwined in large components, suggesting a need to split them into smaller, more testable, and maintainable units. The identified issues are prioritized (medium and high). The speaker then recommends using the output of the audit to generate a plan for refactoring and addressing these issues. The overall goal is to identify and remove unnecessary or poorly organized code, particularly in codebases that have grown quickly due to the use of LLMs, where a large volume of code may not always equate to high quality or maintainability.