| name | quiz |
| description | Generate and grade a graduate-level problem set on a topic or module, with hidden rubric and spaced-repetition follow-up. Use for "/quiz algorithms", "test me on consensus protocols", or timed mock-exam practice. |
Quiz / Problem Set
Generate a graduate-level problem set on the topic in the arguments and grade the user's attempts. If no topic is given, ask — or offer a mixed-domain set drawn from the sections in assessments/breadth-exam.md.
Generating
- Default set: 5 problems — 2 conceptual, 2 quantitative/proof, 1 open-ended design or research-style question. Honor explicit requests for different sizes or formats.
- Pull difficulty from the relevant module in
curriculum/core/: problems should require the module's mastery-checklist skills, not recall.
- Write a rubric per problem before presenting the set, but keep it hidden until grading.
- If the user asks for a timed set, state the time budget per problem up front (use breadth-exam pacing: ~20–30 min per problem).
- If an error log exists in
research/, include at least one variant of a previously missed problem that is due for re-attempt.
Grading
- Wait for the user's attempt. Release hints only on request, smallest first; each hint caps the problem score (note the cap when giving the hint).
- Grade each problem: correctness (40), depth (25), rigor (20), clarity (15). Show the rubric now, quote the specific gap, and state the minimum change for full marks.
- Distinguish slips (knew it, botched execution) from gaps (missing concept). Gaps get logged; slips get a quick re-drill.
After grading
- Report the set score against the program's mastery threshold (≥85% across 3 spaced attempts, per
milestones/roadmap.md).
- Append misses to the error log under
research/ with a +48h re-attempt date.
- If the user has passed three spaced sets on this topic at ≥85%, say so explicitly — that's the signal to advance per the roadmap exit criteria.
