What problem does it solve? When the same AI model writes code and reviews its own work, it carries identical assumptions into both steps, creating systematic blind spots where bugs survive review. This Skill replaces unreliable AI self-review with automated regression tests that mechanically catch recurring bugs like sandbox/production path mismatches and missing SELECT columns. ## Core Features & Use Cases - Sandbox-Mode API Testing: Configure Vitest with Next.js App Router to test API routes without a database by forcing sandbox/mock mode via environment variables. - Bug-Driven Test Strategy: Write tests only for bugs that were actually found, naming them after the bug (e.g., "BUG-R1 regression") so the same defect can never reoccur. - Bug-Check Workflow Integration: Define a custom command that runs tests and builds before any AI code review, then proposes a regression test for every fix. - Use Case: After an AI agent fixes a missing notification_settings field in a profile API for the fourth time, write a contract test asserting all required response fields exist in sandbox mode, then run it as the mandatory first step of every bug-check. ## Quick Start Ask the AI to set up sandbox-mode Vitest regression tests for the API route where a bug was just fixed, and integrate them into the bug-check workflow.