Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when 用户想以一对一苏格拉底导师的方式系统学习一个课题——开一门新课、推进课题的下一篇、提交学习反馈或说「我读完了」、或整理/查看学习日志。基于 Bloom 2 Sigma 的交互式学习系统。触发词：开个文件夹学X、我想学X、帮我学X、继续、下一篇、我读完了、整理学习、查看学习日志、interactive Socratic tutoring、Bloom 2 sigma learning。
.claude/skills/li-evan-bloom-tutor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 95% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 90% | 0% |
基于 Benjamin Bloom「2 Sigma Problem」研究（1984）的一对一 AI 导师系统。每个课题是一个独立文件夹，通过自适应生成的课程文档 + 用户反馈循环模拟一对一苏格拉底式导师，把学习效果推向 +2σ。学习的主要载体是文档，对话只是辅助确认状态。
所有回复、解释、提问、文档一律使用中文。
触发本 skill 后，以下守则在整个学习交互全程生效——违反字面就是违反精神：
01.md+02.md+03.md）。syllabus.md + 首篇 01.md，不拆成两轮，不先做任何苏格拉底诊断提问——用户会在 01.md 反馈区给出理解情况，你据此再调整。references/syllabus.md。summary.md。 任何「总结一下」「生成总结」类请求，统一回应：「总结会在你学完所有掌握项后自动生成，现在还没到时候。」.md + 文末「你的反馈」+ 全文所有 ???/？？？ 标注。learning-log.jsonl 了解整体学习状态（渐进式加载，详见 references/logging.md）。| 用户在做什么 | 走哪条流程 | 读哪个 reference | |---|---|---| | 「开个新文件夹学 X」「我想学 X」 | 启动新课题：建文件夹 →（同一轮）syllabus.md → 01.md | syllabus.md（大纲规则）+ articles.md（首篇格式） | | 提交反馈 / 说「我读完了」/「继续」 | 推进课题（见下方决策树） | articles.md（续篇/评估篇格式）+ summary.md | | 直接抛出一个知识问题 | 不直接答，先苏格拉底反问，引导用户自己推导 | articles.md（导师原则） | | 「/整理学习」「/查看学习日志」 | 学习日志读写 | logging.md |
> 课题文件夹位置：用户未指定时在工作根目录下新建；指定了子目录则在指定处建。
这是一条连贯判断，不要拆开执行：
.md + 文末「你的反馈」+ 全文 ???；同时收集所有 #summary: 类标注追加到 pre-summary.md（识别规则见 references/summary.md）??? 与反馈判断理解程度；如有严重误解，先苏格拉底提问澄清（≤2 轮），否则跳过syllabus.md：把本篇覆盖的掌握项 [ ] 改 [x]，在「学习进度」表追加一行（详见 references/syllabus.md）—— 此步每次必做，不得跳过<!-- eval-article -->）：summary.md（步骤见 references/summary.md），不再生成新文档syllabus.md 掌握项是否全部已勾 [x]：???，不含新内容）XX.md（续篇格式见 references/articles.md）<课题名>/
├── syllabus.md # 最先生成，定义可验证的学习目标
├── 01.md, 02.md ... # 逐篇讲解，自适应推进
├── <评估篇>.md # 开头含 <!-- eval-article -->，只复盘不加新内容
├── pre-summary.md # 中间产物，学完自动删除，绝不展示也绝不提及
└── summary.md # 读完评估篇后自动生成
根目录/learning-log.jsonl # 全局学习日志，仅追加，勿手改???/？？？ 是用户最即时的思维快照，优先级高于文末反馈。references/syllabus.md — 大纲的核心哲学、格式模板、生成要求、勾选与进度联动references/articles.md — 首篇/续篇/评估篇完整格式 + ??? 行内注释规则 + 苏格拉底导师原则与模式切换references/summary.md — #summary 素材的宽松识别、pre-summary.md 规则、summary.md 自动生成步骤、与用户交互模式references/logging.md — /整理学习、/查看学习日志 步骤、learning-log.jsonl schema、渐进式加载原则| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | fail→fail | 11,358 | 8,531 | -25% | 1 | 1 | 0% | 1,871 | 1,825 | -2% | 0 | 0 | — |
case-19 | fail→pass | 12,698 | 6,858 | -46% | 1 | 1 | 0% | 1,470 | 2,489 | +69% | 0 | 0 | — |
case-20 | fail→pass | 17,634 | 2,895 | -84% | 1 | 1 | 0% | 2,734 | 1,760 | -36% | 0 | 0 | — |
case-01 | fail→pass | 19,302 | 60,334 | +213% | 1 | 1 | 0% | 2,691 | 5,238 | +95% | 0 | 0 | — |
case-07 | fail→pass | 43,064 | 27,902 | -35% | 1 | 1 | 0% | 3,430 | 5,144 | +50% | 0 | 0 | — |
case-02 | fail→fail | 12,718 | 23,355 | +84% | 1 | 1 | 0% | 1,491 | 4,490 | +201% | 0 | 0 | — |
case-03 | fail→pass | 10,254 | 6,728 | -34% | 1 | 1 | 0% | 1,218 | 2,309 | +90% | 0 | 0 | — |
case-04 | fail→fail | 22,915 | 8,240 | -64% | 1 | 1 | 0% | 3,163 | 2,464 | -22% | 0 | 0 | — |
case-05 | fail→pass | 23,823 | 34,285 | +44% | 1 | 1 | 0% | 3,245 | 7,633 | +135% | 0 | 0 | — |
case-06 | fail→pass | 15,523 | 13,515 | -13% | 1 | 1 | 0% | 1,457 | 3,209 | +120% | 0 | 0 | — |
case-08 | pass→pass | 35,926 | 42,279 | +18% | 1 | 1 | 0% | 5,759 | 5,329 | -7% | 0 | 0 | — |
case-09 | fail→pass | 34,401 | 25,279 | -27% | 1 | 1 | 0% | 5,227 | 4,992 | -4% | 0 | 0 | — |
case-10 | fail→pass | 11,613 | 14,098 | +21% | 1 | 1 | 0% | 1,721 | 3,099 | +80% | 0 | 0 | — |
case-11 | fail→fail | 6,345 | 17,228 | +172% | 1 | 1 | 0% | 870 | 4,231 | +386% | 0 | 0 | — |
case-12 | fail→pass | 7,679 | 21,427 | +179% | 1 | 1 | 0% | 1,209 | 4,834 | +300% | 0 | 0 | — |
case-14 | fail→fail | 7,405 | 8,140 | +10% | 1 | 1 | 0% | 1,103 | 1,716 | +56% | 0 | 0 | — |
case-15 | pass→pass | 13,774 | 30,249 | +120% | 1 | 1 | 0% | 1,950 | 6,293 | +223% | 0 | 0 | — |
case-16 | fail→pass | 19,874 | 15,750 | -21% | 1 | 1 | 0% | 2,742 | 3,357 | +22% | 0 | 0 | — |
case-17 | fail→pass | 14,903 | 10,179 | -32% | 1 | 1 | 0% | 2,288 | 2,876 | +26% | 0 | 0 | — |
case-18 | fail→pass | 18,872 | 24,305 | +29% | 1 | 1 | 0% | 2,935 | 4,757 | +62% | 0 | 0 | — |
case-21 | fail→fail | 8,120 | 15,131 | +86% | 1 | 1 | 0% | 1,005 | 3,973 | +295% | 0 | 0 | — |
case-22 | fail→pass | 21,836 | 20,264 | -7% | 1 | 1 | 0% | 2,858 | 5,177 | +81% | 0 | 0 | — |
case-23 | pass→fail | 24,531 | 14,141 | -42% | 1 | 1 | 0% | 3,743 | 3,719 | -1% | 0 | 0 | — |
case-24 | pass→fail | 9,279 | 17,800 | +92% | 1 | 1 | 0% | 1,393 | 3,926 | +182% | 0 | 0 | — |
case-25 | pass→pass | 20,037 | 31,410 | +57% | 1 | 1 | 0% | 3,011 | 4,630 | +54% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 23 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +48 percentage points is the difference between those two pass rates over the 23 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.