quiz

对用户进行主题测验、按深度评分并记录 AI 置信度与凭证。

针对某个主题出题并评估用户答案的深度，把回答按深度 rubric 分类，同时记录 ai_confidence 和证明凭证。用户要求测验、评估知识边界或想知道掌握程度时用，它不会无凭空猜级别，只有证据才给分。结果能保留可复核的记录，方便后续跟踪学习效果和信度。

# /quiz — assessment at the evidence boundary

The only workflow that produces ai_confidence. Never guess a level without answers.

**First-touch (run this first):** `.venv/bin/python -m brain first-touch quiz`. If it
prints a paragraph, this is the user's first quiz — open your reply with that text
verbatim, then continue. If it prints nothing, skip it and proceed.

1. Pick the target: his choice, or the highest-value candidate from
   `.venv/bin/python -m brain gaps` (weak-but-important beats unknown-and-optional).
2. Read the evidence: `brain search "<topic>" -k 5`, then the full matching notes.
   Read `rubrics/depth.yaml`. Establish the currently evidenced level and what's thin.
3. Gene