Tdd

TDD Reference

Practical guidance for test-driven development: writing good tests, designing testable interfaces, mocking correctly, and refactoring safely.

Scope Constraints

• Read-only reference skill — provides guidance, does not modify code
• Does not run tests, generate test files, or configure test runners
• Complements superpowers:test-driven-development workflow skill

Inputs

• Context (required): What the user is testing or designing
• Phase (optional): red (writing failing test), green (making it pass), refactor (cleaning up)

Input Sanitization

• Context descriptions: free text, reject null bytes.

Procedure

1. Identify the TDD phase. If unclear, ask: "Are you writing a new test, making one pass, or refactoring?"
2. Load the reference file matching the phase or question:
   • Writing tests → references/tests.md
   • Designing interfaces for testability → references/interface-design.md
   • Mocking decisions → references/mocking.md
   • Refactoring after green → references/refactoring.md
   • Module depth and API design → references/deep-modules.md
3. Apply the reference guidance to the user's specific code context.
4. Flag anti-patterns from the loaded reference when found in user code.

Output Format

## TDD Guidance: [phase/topic]

[Specific advice applied to user's code context]

  Pattern: [recommended approach]
  Anti-pattern: [what to avoid and why]

Reference: [which reference file was consulted]

Gotchas

• Tests that assert implementation details (e.g., checking internal method calls) break on every refactor — test behavior and outputs, not how the code does it
• Mocking at the wrong boundary (e.g., mocking the function under test, or mocking too deep) — mock at system boundaries (HTTP, DB, filesystem), not internal modules
• Assertion-free "tests" that just call code without checking results — every test needs at least one meaningful assertion
• beforeAll shared state between tests causes order-dependent failures — prefer beforeEach for fresh state per test
• Testing private methods directly indicates poor interface design — if you need to test it, it should be part of the public API
• Snapshot tests that get --updated without review become rubber stamps — review every snapshot diff carefully

// WRONG: testing implementation details
expect(service.internalCache.size).toBe(3);
// RIGHT: testing behavior
expect(await service.getUser("abc")).toEqual({ id: "abc", name: "Alice" });

// WRONG: mocking the thing you're testing
vi.spyOn(calculator, 'add').mockReturnValue(5);
expect(calculator.add(2, 3)).toBe(5); // tests nothing
// RIGHT: mock dependencies, test the unit
vi.spyOn(httpClient, 'get').mockResolvedValue({ rate: 1.5 });
expect(await converter.convert(100, 'USD', 'EUR')).toBe(150);

References