Restore architecture in agent-written code
Over a quarter, agents have filled out a service where each feature lives in its own module — and each module is fine, but they leak: business logic seeps into HTTP controllers, the DB schema is queried directly from three layers, and there's no clear domain boundary anymore because no single agent run ever 'owned' the architecture. The system works and ships. As the staff engineer, how do you reassert clean abstraction boundaries across a largely machine-written codebase, and what do you change about HOW you delegate so the boundaries hold next quarter?
find_layer_violations(from_paths: list[str], to_paths: list[str]) → list[str][["controllers/orders.ts","services/orders.ts","controllers/orders.ts"],["services/orders.ts","repositories/orderRepo.ts","repositories/orderRepo.ts"]]out["controllers/orders.ts -> repositories/orderRepo.ts: skips-layer"][["repositories/db.ts","services/pricing.ts"],["controllers/checkout.ts","controllers/checkout.ts"]]out["repositories/db.ts -> controllers/checkout.ts: upward","services/pricing.ts -> controllers/checkout.ts: upward"][["controllers/health.ts","lib/logger.ts"],["lib/logger.ts","services/orders.ts"]]out["lib/logger.ts -> services/orders.ts: shared-reaches-in"]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.