What problem does it solve? Code written before tests produces tests that pass immediately and prove nothing, and prompt-based baseline scenarios can silently fail for the wrong reason when the agent already knows the lesson or the prompt gives away the answer. This Skill enforces a strict test-first workflow and ships a checker that detects both leaks before a baseline run is dispatched. ## Core Features & Use Cases - Red-Green-Refactor Discipline: Mandates writing one failing test, verifying it fails for the right reason, then writing minimal code to pass, with checklists and rationalization rebuttals. - Scenario RED Auditing: The bundled redcheck.py script checks a scenario prompt against the agent's inherited context (CLAUDE.md cascade, memory facts) for prior coverage and against telegraphing markers, with exit codes 0-3 for gating. - Testing Anti-Pattern Reference: A companion reference covers testing mock behavior, test-only production methods, incomplete mocks, and mocking without understanding dependencies. - Use Case: Before dispatching a subagent with a prompt meant to prove it takes the wrong action, run redcheck.py with --corpus-cascade on the dispatch directory to confirm the lesson is not already in its always-loaded context. ## Quick Start Ask the agent to implement a bug fix using test-driven development, writing and running a failing test before any production code.
