| name | tdd |
| description | Guides execute-inspect-adjust development and t-wada Red-Green-Refactor TDD. Use for exploratory implementation or when stable behaviour needs an executable test-first contract. |
Use TDD when a behaviour should become an executable contract. For unfamiliar APIs, prototypes, and data exploration, use execute-inspect-adjust first; not every probe needs a test.
Before writing a test, identify the public interface and the seam where its behaviour can be observed. When CONTEXT.md or relevant ADRs exist, read them so test names and interfaces use the project's vocabulary.
For test design and anti-patterns, read references/testing.md. For test doubles and system boundaries, read references/mocking.md.
Cycle
- Sketch behaviours as placeholder tests, identify their public seams, and choose one slice. Start bug fixes with a regression test.
- Red — Write the smallest failing test and confirm the expected failure.
- Green — Write the minimum production code that passes it.
- Refactor — Improve test and production code while keeping tests green.
- Run affected tests after each green and refactor step, then repeat.
Keep each test focused on one observable behaviour; do not weaken tests to make the build pass.
Rules
- Test through a public interface at an agreed seam, not through private methods or implementation details.
- Work in vertical slices: one seam, one test, and one minimal implementation at a time. Do not write all tests first and all production code afterwards.
- Avoid implementation-coupled and tautological tests; expected values should come from an independent source of truth.
- Mock only system boundaries. For HTTP APIs, prefer request-level interception such as MSW over mocking
fetch, an API client, or an internal module.
For runner-specific syntax, read the matching file in references/.
