What problem does it solve?
When AI models write and then review their own code, they carry identical assumptions through both steps, creating systematic blind spots that lead to recurring, hard-to-detect regressions — most commonly sandbox/production path mismatches that break functionality in one environment but not the other.
Core Features & Use Cases
- Sandbox-mode API testing: Run fast, database-free API tests by forcing sandbox mode, no need for local database setup.
- Automated bug-check workflows: Integrate test runs as the first step of post-change code reviews to catch bugs mechanically before AI review.
- AI blind spot pattern library: Pre-built test templates for the most common AI-introduced regressions, including SELECT clause omissions, error state leakage, and missing rollback logic.
- Use Case: After an AI agent modifies a Next.js API route, use this skill to write a regression test that ensures the response shape matches expectations in both sandbox and production modes, preventing the bug from being re-introduced.
Quick Start
Use the ai-regression-testing skill to run automated regression tests for recently modified backend API routes to catch AI-introduced blind spot bugs.