tdd-workflow

Write a failing acceptance test from the bug description first, then a failing unit test named with the bug ID. Implement the minimal fix to turn tests green, verify coverage, and record the root cause in the memory database for traceability.

RED writes a failing test proving the target behavior is missing, GREEN adds the minimal code to pass it, and REFACTOR improves code quality without changing behavior. Each phase gets its own commit, and CI must show red then green.

A minimum of 80 percent combined coverage across unit, integration, and end-to-end tests. Edge cases, error scenarios, and boundary conditions must all be covered, not just happy paths.

Test code should follow DAMP (Descriptive And Meaningful Phrases) over DRY. Keeping setup inline and slightly repetitive makes each test readable on its own, while over-abstracted helpers hide critical setup details.

Avoid testing internal implementation details instead of visible behavior, using brittle CSS selectors, creating dependencies between tests, writing code before tests, and skipping error paths. Each test should be independent with descriptive names.

Test failure fixes are limited to three rounds. If the issue persists beyond that, the workflow triggers context compression or a fresh root cause analysis instead of continued trial-and-error patching.