The QA tool embedded in the app (task QA-UXGRID1). Three layers: documentation explains the process (static-ish), configuration defines what we test (slow-moving, git change-controlled JSON), results shows testing happening (live).
How and why we test — the process itself. Evolves slowly, by discussion.
What we test: dimensions (the axes), test classes and their derived instances, the canary catalogue, decisions. Designing tests and fixing bugs updates this layer — every change is a git commit.
Testing happening: the retest grid (commits across, cases down — same idea as the backend's kflow-admin), the manual-run recorder, recent runs.
Data files live in qa/data/ (source of truth, git-tracked). These pages only display them.
Manual test runs are recorded as test:start/test:verdict kflow checkpoints — the same
convention the backend's katrun robot uses, so all testing speaks one language.