Chapter 12: Testing AI-Generated Code — A Practical Framework

The Testing Paradox of Vibe Coding

Vibe coding's greatest strength — speed — creates a dangerous gap. You can build features in hours that would normally take days. But without tests, you are building a house of cards. Every new feature risks breaking existing ones. Bugs reach users. Debug sessions replace shipping sessions.

This chapter is not about test-driven development or 100% code coverage. It is about a pragmatic, AI-assisted testing strategy that gives you meaningful confidence without slowing down your velocity.

---

The Testing Pyramid for Vibe Coders

Classic software engineering recommends many unit tests, fewer integration tests, and very few end-to-end tests. For vibe-coded projects, I recommend inverting the traditional emphasis:

Priority 1: Critical Path E2E Tests The flows that make you money or break your product. Sign up → subscribe → use core feature. These must always work.

Priority 2: API Route Integration Tests Your API routes are the contract between your UI and your data. Test inputs, outputs, and error cases.

Priority 3: Unit Tests for Business Logic Complex calculations, data transformations, and validation rules. These are fast, reliable, and high-value.

---

Generating Tests With AI

The best prompt for generating tests is one that provides the full function context and explicitly lists the cases you want covered.

Unit Test Generation Prompt:

Write Jest unit tests for this utility function. Cover:
1. The happy path with valid input
2. Each validation rule being violated
3. Edge cases: empty string, null, undefined, very long input, special characters
4. The exact error messages returned for each failure

Function:

Use Jest with TypeScript. Mock any external dependencies (database calls, API calls). 
Output the complete test file.

API Route Test Prompt:

Write integration tests for this Next.js API route using Jest and node-mocks-http.

Test these scenarios:
1. Valid request — correct inputs, authenticated user — returns 200 with expected data
2. Unauthenticated request — returns 401
3. Missing required fields — returns 400 with validation error
4. Valid request but user has no credits — returns 403
5. Database error simulation — returns 500 with appropriate error message

Mock:
- Supabase client (return mock session/data)
- Any external API calls

Route code:

---

What to Always Test (The Non-Negotiables)

These categories of tests should exist in every vibe-coded project:

Auth boundary tests: Every authenticated route should have a test that verifies an unauthenticated request is rejected with a 401.

Credit/quota enforcement: Any feature that charges credits or enforces usage limits needs a test that verifies the limit is actually enforced.

Payment webhook handling: Your Stripe webhook handler needs tests for each event type. A broken webhook means users pay and get no credits.

Data validation: Any user-submitted data going into your database should have validation tests that verify malformed data is rejected before it reaches the DB.

---

The AI Debugging Loop

When a test fails and the cause is not obvious, use this debugging prompt:

This test is failing with this error:
[paste error]

Here is the test:
[paste test]

Here is the code being tested:
[paste code]

Diagnose: Is this a bug in the test, a bug in the code, or an incorrect mock setup?
Provide the fix for whichever is wrong.

---

Integrating Tests Into Your Sprint

The most common failure mode is intending to "write tests later" and never doing it. Instead, insert testing into your natural workflow:

• When you finish an API route → immediately ask AI to generate the integration tests
• When you finish a utility function → immediately ask AI to generate unit tests
• Before every deploy → run the test suite. Do not ship with failing tests.