What problem does it solve? It prevents developers and AI agents from writing implementation code before a failing test exists, eliminating test-after-the-fact habits and weak test suites that pass even when the code is broken. ## Core Features & Use Cases - Phased TDD Cycle: Defines the Plan → Red → Green → Mutate → Refactor workflow with explicit phase signals ([RED], [GREEN], [MUTATE], [REFACTOR]) so progress is auditable. - Mutation Testing Gate: After Green, weaken or delete the guard under test and confirm the suite goes red; optionally run gremlins (Go) or Stryker (TS/JS) to catch tests that are theater. - Subagent Isolation: When diffs grow large, split Red (test writing) and Green (implementation) into separate subagent contexts to avoid same-author blind spots. - Use Case: You start a new feature and the tdd-guard.sh hook blocks your edit with "No test file found" — load this protocol, write the failing test first, then implement. ## Quick Start Ask the AI to load the tdd-protocol skill and guide the current task through the Red, Green, Mutate, and Refactor phases starting with a failing test.