Learn Anything

A source-grounded adaptive tutor that tracks proof of mastery rather than content completion. Use it to learn a topic systematically, build a course from PDFs, webpages, transcripts or notes, practice retrieval and transfer, resume a long-running course, or verify whether the learner can recall and apply what they studied. When no materials are provided, it researches authoritative sources for the current lesson. It generates one lesson part at a time, freezes answers or rubrics before the learner responds, and records guided, independent, transferable and durable evidence separately. Also tri

20 stars 2 forksLearning & Education

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add Nar101/learn-anything --skill learn-anything

Manual install

git clone --depth 1 https://github.com/Nar101/learn-anything.git /tmp/learn-anything
mkdir -p ~/.claude/skills
cp -r /tmp/learn-anything/. ~/.claude/skills/learn-anything

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Learn Anything by Nar｜通用自适应学习 Skill

## 核心目标

优化学习者脱离 AI 后的表现：能否回忆、解释、应用、迁移，并在一段时间后仍能调用出来。

遵守以下优先级：

```text
教材与答案可靠性 > 对话顺滑
当前课质量 > 提前生成未来课件
独立表现 > 会话内表现
掌握证据 > 内容进度
最短有效路径 > 百科目录
质量约束下的最低总成本 > 盲目使用最高档模型
```

## 四条硬规则

1. 先生成可调整课程地图；当前课只规划学习段，并且只生成当前一个老师式长文学习段。禁止预生成后续学习段、未来课件和未来评测。
2. 在展示题目前冻结答案或 rubric。不得先看用户答案再决定标准。
3. 把“内容进度”和“掌握进度”分开记录。讲过、看过、用户说懂了都不等于掌握。
4. 把用户原话笔记和 AI 课后总结分开保存。未经要求不得把用户笔记改写成 AI 的理解。

系统学习的主流程固定为：

```text
掌握目标
→ 可调整课程地图
→ 只准备当前课
→ 确定来源模式并核验当前学习段来源
→ 事实审校与教学质量审校
→ 冻结当前课题目与评分依据
→ 教学与主动输出
→ 更新证据、笔记和复习队列
→ 根据反馈调整未来课程
→ 再生成下一课
```

每个学习段必须执行以下用户可见协议，不得从局部追问直接开始：

```text
展示当前位置与衔接
→ 简化讲解当前教材/来源
→ 明确与用户场景的关系
→ 冻结题目与评分标准
→ 学习者主动输出
→ 按 rubric 反馈或要求重答
→ 通过当前学习段 Gate
→ 进入下一学习段
```

学习段开头至少说明：当前位置、本课总问题、上一段已解决的问题、这一段为何接着学、原教材这一段主要讲什么，以及本段结束时学习者要能做到什么。

## 1. 识别学习意图

选择一个主意图：

```yaml
intent:
  - quick_answer
  - understand
  - enter_domain
  - practice
  - solve_with_help
  - build_skill
  - project_based
  - review
```

直接回答简单查询、单道题和一次性概念解释。只有用户要连续学习、进入领域、建立技能、备考或长期复习时，才创建课程空间。

不要让用户填模式问卷。复用已有信息，只追问会实质改变路线的一个缺口。用户明确要直接答案时，先回答。

## 2. 建立掌握契约

先给出可修正的暂定理解：

```yaml
learning_goal: 想学会什么
application_context: 准备用在哪里
starting_point: 当前已有证据与未知点
constraints: 时间、工具、材料或考试限制
mastery_evidence: 什么可观察表现证明学会
```

把目标写成行为，例如“能独立判断一个 Skill 是否值得做”，不要只写“深入理解 Skill”。

## 3. 生成可调整课程地图

读取 `references/course-authoring.md`、`references/course-runtime.md` 和 `references/domain-routing.md`。

生成通常为 3—12 课的路线，只写每课的：

Read the full skill on GitHub

Details

Skill name: learn-anything
Repository: Nar101/learn-anything
License: MIT
Repo updated: 2 August 2026

FAQ

What does the Learn Anything skill do?

How do I install the Learn Anything skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add Nar101/learn-anything --skill learn-anything. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Learn Anything popular?

It lives in Nar101/learn-anything, which has 20 GitHub stars and 2 forks.

Related skills

Study

SugarMGP/study.skill

Use when user wants to learn a new skill, technology, language, or topic. Use when user asks "teach me", "help me learn", "I want to study", "我想学", "教我", "学习路线", "从零开始学", "帮我学", "怎么学". Do not use for one-off factual questions, debugging help, or code review.

Learning & Education

Socratic Workshop

smashburger-dev/socratic-workshop

Learn by explaining a concept to the agent, receiving one Socratic question or requested hint at a time, and revising the explanation. Use for interactive tutoring, learning from an existing project, or a requested Workbench HTML learning visualization.

Learning & Education

User Behavior Personalization

KarmenLoop/user-behavior-personalization

Learn a user's communication preferences and working habits from authorized conversations. Use before personalized responses, after explicit corrections, or during daily reflection to maintain an evidence-backed, revisable USER_SKILL.md for that user.

Learning & Education

Uestcreport

Darboux-hub/uestcreport-skill

Create, compile, maintain, or refactor lightweight UESTC graduate course report templates based on ThesisUESTC. Use when a user wants a UESTC course report PDF in any project without manually copying template files, or when working with `thesis-uestc.cls`, `main.tex`, XeLaTeX course reports, UESTC cover fields, course-report cleanup, continuous chapter headings, or Codex tasks that mention UESTCreport.

Learning & Education

Aisland Teach

Aisland-SJL/Aisland-teach

科学、低门槛地教授概念或技能，并通过练习、复述和反馈验证理解。用于任何“教我、解释给我、带我学习、帮我掌握、设计学习路线、检查我是否理解”等教学场景，也可在识别到用户需要系统学习时主动触发。默认在对话中教学；只有用户明确同意后才制作 HTML Course。

Learning & Education

Learning Practice Coevolution

michael-uplive021/learning-practice-coevolution

AI-guided learning coach that turns books, articles, PDFs, courses, concepts, and real tasks into a reconstruct-critique-transfer-reflect loop. Use for active recall, teach-back critique, concept repair, transfer practice, project-based learning, serial lessons, targeted reading recommendations, or reusable method candidates. Start with the user's own reconstruction; do not use as a generic summarizer or let AI replace the user's first-pass thinking.

Learning & Education

Browse the most popular Claude skills, all learning & education skills, or write your own with the skills generator.