Enforce strict red-green-refactor: generate failing tests from a spec, observe them fail, then permit implementation. Use for: TDD workflow, test-first, red-green-refactor, must see red, failing test first, generate tests before code, bug fix TDD, regression test before fix. Do not use for: writing tests after code exists (use test-generator) or auditing whether tests are load-bearing (use test-effectiveness-reviewer).