/quiz [<course-slug>] [--lessons 1,3] [--weak] [--n <questions>]

Interactive retrieval practice — the core of the testing effect. Read docs/pedagogy.md (quiz rules) and docs/vault-schema.md (mastery update rule, quiz-log schema) first. Vault: <vault>.

Scope selection

• --lessons: exactly those lessons of the course.
• --weak: lessons with mastery < 60, plus stale ones (last_quizzed > 21 days, mastery < 85).
• Default: interleaved mix across the course's status: read|quizzed|mastered lessons, weighted toward low mastery and staleness. No course given → mix across all active courses (max interleaving).
• Read the scoped lessons AND their concept notes to ground questions. Default length: 6–10 questions (--n overrides).

Session loop (one question at a time)

1. Before asking, privately write a 10-point rubric for the question.
2. Ask conversationally. Mix formats per pedagogy.md: free recall, explain-why, worked problems (user answers in text/LaTeX-ish), transfer questions. Never multiple choice.
3. Every 2–3 questions, ask for a confidence rating (1–5) before grading — flag high-confidence misses.
4. Grade /10 against the rubric with partial credit. Give elaborative feedback: correct answer, why, misconception note if relevant, link to the concept note. If the missed knowledge is encoded in the user's mind palace (grep palace/ for a link to the concept), add the location as a retrieval cue ("this lives in your palace — <room>, <spot>") — cue the place, never recite their image. Then next question.
5. The user can stop anytime ("stop", or the equivalent in their language) — finish the session bookkeeping with what was answered.

After the last question

All rules below are defined in docs/vault-schema.md (lesson type) — apply them exactly.

1. Overall score /100 (weighted by rubric points), plus a lesson score (0–100) for each covered lesson from its own questions only.
2. Write the quiz log per templates/quiz-log.md at courses/<slug>/quizzes/YYYY-MM-DD Quiz <scope>.md — per-question record, weak_concepts (anything graded ≤ 5, plus high-confidence misses), and the per-lesson ## Result table. A session spanning several courses writes one log per course. If the filename already exists, append (2), (3), … — never overwrite a log.
3. Update each covered lesson's frontmatter, reading last_quizzed before overwriting it: mastery (first quiz, i.e. last_quizzed empty → lesson score; else round(0.4×old + 0.6×lesson score)), then last_quizzed = today, then status (mastered iff mastery ≥ 85 and the last two lesson scores — this one and the previous quiz log's — are ≥ 80, else quizzed; this also demotes a slipped mastered). Recompute each affected course's lessons_done. Set each involved concept note's mastery to 10 × its most recent question grade.
4. For every failed item (≤ 5/10): append 1–2 new targeted cards to the relevant lesson's card file (or concepts/Inbox Cards.md, deck #flashcards/inbox, if courseless), attacking exactly the gap. New cards have no SR comment → due immediately. Never touch existing card lines with <!--SR: comments; bump card_count.
5. Report: score, per-lesson mastery changes, cards spawned, and what /study should prioritize tomorrow.