ZeKaiNie/universal-examprep-skill / full
universal-exam-cram-coach-full
Verified帮助学生在临考前进行结构化极速复习:解析课程资料/大纲/重点,按章节生成 wiki 知识库与标准题库,组织针对性刷题与判分,并记录复习进度和错题。当用户即将考试、需要快速复习计划、练习题、错题复盘或考前小抄时使用(关键词:期末/备考/复习/刷题/划重点/错题;exam, cram, study plan, quiz, review)。不适用于长期学习规划、与考试无关的写作或编程任务。
benchmark/items/README.md
View on GitHub# items/ —— 金标题集(ground truth)
[English](README.en.md) · 中文
测试的核心。每道题是一行 JSON(JSONL 格式),既给"标准答案",也给"答案出自材料哪一句"
(supporting span)——后者让裁判判"答得是否忠于材料",也是数值题判分的依据。
复制 `items.example.jsonl` 为 `items.jsonl`,按下面规范、依据你**自己的课件/作业**改写。
## 字段规范
| 字段 | 必填 | 说明 |
| :-- | :-: | :-- |
| `id` | ✓ | 唯一编号,如 `q1` |
| `question` | ✓ | 题面(学生真会问的问题) |
| `gold_answer` | ✓ | 标准答案(简洁、确定)。越界题(answerable=false)留空 `""` |
| `supporting_span` | ✓* | 支撑答案的**材料原文摘抄** + 出处;越界题留空 |
| `source_file` | ✓* | 该题出自哪个材料文件,如 `materials/ds/ch5_search.pdf` |
| `answer_type` | ✓ | `factual`(事实)/ `definition`(定义)/ `numeric`(计算/数值) |
| `answerable` | ✓ | `true`=材料能答;`false`=**越界探针**(材料里没有,正确行为是弃答) |
| `tolerance` | numeric 时 | 数值题允许误差,如 `0`(精确)或 `0.01` |
## 怎么编(建议流程)
1. **通读你的讲义/作业**,挑出 15~40 个高频考点,写成问题 + 标准答案。
2. **每条标注 `supporting_span`**:把答案依据的那句材料原文抄进来,并写清 `source_file`。
(可以让 AI 起草,但你要逐条核对——AI 起草、人工把关。)
3. **故意放几条越界探针**(`answerable:false`):学生可能会问、但你的材料根本没讲的问题
(如"考试在哪个教室""老师叫什么")。这是检验 skill"宁可说不知道也不编"的关键——
它直接对应 skill 宣称的"物理防幻觉"。**优先出 near-miss 越界题**(话题贴着材料、但材料里没有
确切答案的题,如材料讲了某算法却问它没提的一个参数),而非明显跑题的——明显越界太好答,
near-miss 才分得出"真读了材料并知道边界"与"看着相关就编"。详见 [`docs/judge-calibration.md`](../docs/judge-calibration.md)。
4. **计算题**(`answer_type:"numeric"`):尽量来自作业,记准确数值 + `tolerance`;这类题由脚本
**确定性判分**(不经过 LLM 裁判),最干净。
5. 存成 `items.jsonl`。**小而全标注 > 大而粗糙**:30 题左右、每题都有金标和出处,就足够出一份可信报告。
> 平台/英文版规划:题面与裁判语言无关,现在用中文,将来做英文版时同一套 harness 直接复用,
> 只需另出一份英文 `items.jsonl`。
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |