test-author

What problem does it solve? Green test suites can lie: tests pass while users see broken behavior because facts were asserted at the wrong level, such as checking source strings instead of rendered output. This Skill turns a reviewed spec and architecture into a TEST_MATRIX.md grid and a test suite where every fact is pinned to the level that can actually see it fail. ## Core Features & Use Cases - Test matrix derivation: Projects every spec fact into a checkable grid organized by architecture node, with each row stating both the expected behavior and the regression fence it must never violate. - Level ladder pinning: Assigns each fact to string, DOM-text, browser-computed, or pixel level so visibility, layout, color, and interaction facts are asserted in a real browser engine rather than against source text. - Red-first and traceability enforcement: Requires every new test to fail against the pre-change state before implementation, pins the expected skip-set, and keeps a standing traceability test that fails on missing rows, duplicate ids, or uncovered spec facts. - Use Case: A spec says a warning banner turns red when audio clips. Instead of a string check on the template, the fact is pinned at browser-computed level, the test opens a real browser, triggers a clip, and reads the computed color, failing red before the fix ships. ## Quick Start Ask the agent to derive the test matrix and pin test levels from your reviewed spec and architecture documents.