E2E (future Playwright) — critical journeys: complete study session, full quiz
Component (RTL)         — interactive UI: answer submission, feedback display
Unit (Vitest)           — pure logic: answer-check, fsrs, quiz-engine, danish-input

Coverage audit

npm test -- --run --coverage 2>&1 | tail -40
find src -name "*.test.*" | sort

Area	File	Status
Answer checking	`src/lib/answer-check.ts`	20 tests
FSRS scheduling	`src/lib/fsrs.ts`	missing
Quiz engine	`src/lib/quiz-engine.ts`	missing
Danish input	`src/lib/danish-input.ts`	missing
useQuiz hook	`src/hooks/useQuiz.ts`	missing

Test quality rules

Describe behaviour, not implementation: it('accepts ae as ae equivalent') not it('calls normalise()')
Cover: empty input, ae/oe/aa chars, case + whitespace, error paths (Supabase failure, empty array)
BDD naming: describe('when user types wrong answer') { it('shows correct answer in feedback') }
Co-locate test files: ComponentName.test.tsx next to ComponentName.tsx
Regression rule: write a failing test before fixing any bug

Pre-release quality gate

npm test -- --run — all tests pass
Answer checking covered for any new exercise types
All 6 grammar topic slugs resolve to content
FSRS tested: new card (state=0) → state>=1 after Good rating
No console errors on npm run dev
Layout checked at 375px mobile viewport

Output format

## QA Review — <feature>
| Area | Status | Gap |
|------|--------|-----|
| answer-check | 20 tests | — |
| FSRS | 0 tests | HIGH: schedule new card, Again resets |

Missing tests (priority):
1. [HIGH] FSRS new card scheduling
2. [MEDIUM] useQuiz answer submission flow

Verdict: Ready / Merge with todo / Needs tests first