What problem does it solve? When implementing tasks from an OpenSpec change, AI agents tend to batch-write tests, write production code before tests fail, or skip refactoring — producing code that looks tested but lacks real TDD discipline. This Skill enforces an atomic per-test RED-GREEN-REFACTOR cycle so every line of production code traces back to a previously failing test. ## Core Features & Use Cases - Iron Law Enforcement: No production code may be written without first observing a test fail for the right reason, with explicit red-flag phrases that halt violations. - Mandatory Gherkin Coverage: Every scenario in the change's spec.md must become at least one passing acceptance test before the slice can ship. - Per-Test State Machine: Each test (acceptance, unit, edge case, helper, error path) traverses RED, VERIFY-RED, GREEN, VERIFY-GREEN, and a mandatory REFACTOR assessment before the next test begins. - Use Case: While an implementer subagent works through tasks.md for an OpenSpec change, it loads this Skill first, translates each Gherkin scenario into a failing acceptance test, writes minimal code to pass it, assesses refactoring, and only then moves to the next test. ## Quick Start Ask the agent to implement the tasks of an OpenSpec change using strict TDD, covering every Gherkin scenario in spec.md with at least one test.
