What problem does it solve? When implementing a single acceptance criterion or feature unit, tests written after the code become verification rather than specification, and coverage numbers alone cannot prove assertions actually catch bugs. This Skill enforces a disciplined test-first cycle so each unit of behavior change gets a failing test before implementation, a minimal green pass, a refactor, and evidence that every branch is genuinely exercised. ## Core Features & Use Cases - Red/Green/Refactor cycle: Write a failing test first, implement the minimum code to pass, then refactor while staying green, with optional aiwf phase promotion (aiwf promote M-NNNN/AC-<N> --phase red|green|refactor|done) for projects tracking TDD phases. - Red-first diff-shape gate: When a project declares test-path globs, the red-phase promotion mechanically refuses if implementation files are dirty before the test, enforcing test-first ordering. - Hard-rule branch-coverage audit: Walk every reachable conditional branch in the diff and confirm an explicit test exercises each side, including defensive and error-handling paths. - Required vacuity check: Invoke wf-vacuity after the audit to verify covered assertions can actually fail, using mutation testing tools where available. - Use Case: While implementing one acceptance criterion of an in-progress milestone, run this cycle to produce the failing test, minimal implementation, and coverage evidence before the milestone ritual marks the criterion met. ## Quick Start Ask the AI to run the wf-tdd-cycle for the acceptance criterion you are about to implement, starting with a failing test before any implementation code.
