knowledge-learning

name

knowledge-learning

description

Progressive knowledge learning system with assessment-driven teaching. Use when user wants to learn new concepts, saying things like "我想学习X", "教我X", "帮我理解X概念", "学习X知识", or "explain X to me". Supports assessment through questioning, adaptive content delivery based on user's current understanding, interactive Q&A, and optional learning report generation. Features structured output to knowledge/ and learning-process/ directories with index.md tracking.

Knowledge Learning

A structured workflow for teaching new concepts through assessment-driven, progressive learning.

Core Principles

1. Assessment First: Start by understanding what the user already knows
2. Adaptive Teaching: Deliver content matched to user's comprehension level
3. Progressive Disclosure: Begin with fundamentals, advance only when ready
4. Interactive Learning: User-driven questioning and exploration
5. Reflection: Optional learning process documentation

Output Structure and File Management

Directory Structure

All learning outputs are organized in the following structure within the project's base directory:

knowledge-learning/          # Skill基础目录
├── index.md                 # 索引文件：记录所有学习会话的摘要和链接
├── knowledge/               # 知识报告目录
│   ├── [date]-[topic]-knowledge.md
│   └── ...
├── learning-process/        # 学习过程记录目录
│   ├── [date]-[topic]-learning-process.md
│   └── ...
├── SKILL.md                 # Skill定义文件
└── ...                      # 其他Skill相关文件

File Naming Convention

• Knowledge Reports: [YYYY-MM-DD]-[topic-slug]-knowledge.md
  • Example: 2024-01-15-machine-learning-knowledge.md
• Learning Process: [YYYY-MM-DD]-[topic-slug]-learning-process.md
  • Example: 2024-01-15-machine-learning-learning-process.md
• Index File: index.md (fixed name, updated after each learning session)

Path Resolution

When generating files:

1. Use the skill's base directory (knowledge-learning/) as the root
2. Create knowledge/ and learning-process/ subdirectories if they don't exist
3. Always generate files in the appropriate subdirectory with proper naming
4. Update index.md in the skill's base directory after each learning session
5. Use relative links in index.md to reference the generated files

Workflow

Phase 0: Pre-learning Assessment (预学习评估)

Goal: Understand user's learning history and avoid repetition

1. Read index file (读取索引):

   • Check if index.md exists in the skill's base directory (knowledge-learning/)
   • If exists, read the file to understand:
     • Previous learning sessions (日期、主题)
     • Key topics already covered (关键要点)
     • Learning progress patterns (学习进度模式)

2. Analyze learning history (分析学习历史):

   • Identify if the requested topic has been previously learned
   • Check for related topics that might provide foundation
   • Note any recurring themes or interests

3. Determine learning approach (确定学习策略):

   • New topic: Proceed with full assessment (Phase 1)
   • Previously learned topic:
     • Ask if user wants a review or deeper dive
     • Reference previous learning: "我看到你之前在[日期]学习过[主题]，想在那个基础上继续深入吗？"
   • Related topic:
     • Build on previous knowledge: "基于你之前对[相关主题]的理解，我们可以从那里开始扩展。"

4. Update learning context (更新学习上下文):

   • Note the current learning session's starting point
   • Prepare to reference previous learnings when relevant

Phase 1: Initial Assessment (评估阶段)

Goal: Understand user's current knowledge level

1. Ask user to self-explain (第一个问题):

   在我开始教学之前，请你先用自己的话解释一下：你目前对[知识点]的理解是什么？
   你可以说出任何你知道的内容，哪怕只是模糊的印象或片段。
   

2. Analyze user's response to identify:

   • Knowledge gaps (完全不了解的部分)
   • Misconceptions (理解偏差)
   • Existing foundation (已有基础)
   • Learning style indicators (学习风格线索)

3. Generate 4-7 targeted questions based on their self-explanation:

   • Focus on knowledge boundaries (他们理解的边界在哪里)
   • Probe misconceptions (验证可能的误解)
   • Assess prerequisite knowledge (前置知识是否具备)

4. Ask questions ONE AT A TIME:

   • Wait for answer before next question
   • Adapt subsequent questions based on previous answers
   • Keep questions conversational, not exam-like
   • Example format: "你提到了X，那你觉得X和Y之间有什么关系？"

5. Complete assessment after 5-8 questions total (including initial self-explanation)

Phase 2: Adaptive Teaching (教学阶段)

Goal: Deliver foundational knowledge matched to user's level

Based on assessment results, provide:

1. Core concept explanation (核心概念):

   • Start with the simplest, most fundamental aspect
   • Use analogies related to user's existing knowledge
   • Avoid jargon unless user demonstrated familiarity
   • Keep initial explanation to 3-5 paragraphs maximum

2. Scope boundaries (明确范围):

   • Explicitly state what you're NOT covering yet
   • Example: "现在我们先理解基础部分X。关于高级话题Y和Z，我们稍后再讨论。"

3. Concrete examples:

   • Provide 1-2 simple, relatable examples
   • Match examples to user's context if known

4. Comprehension check:

   • End with: "这部分清楚吗？你可以提出任何疑问，或者让我继续讲解下一部分。"

Phase 3: Interactive Exploration (互动阶段)

Goal: User-driven deepening of understanding

1. Wait for user questions - Do not proceed automatically
2. Answer questions:
   • Directly address what they asked
   • Check if answer creates new questions
   • Offer to elaborate or move to related topics
3. Suggest next steps only when user seems ready:
   • "你现在理解了X，想继续了解Y吗？"
   • "还有什么疑问，或者我们可以进入下一个概念？"

Phase 4: Cycle or Conclude

Two paths:

A. Continue Learning (循环):

• Return to Phase 1 for next concept layer
• Or return to Phase 2 for deeper dive
• User signals: "继续", "下一步", "还有什么", "深入讲讲"

B. Generate Learning Report (生成报告):

• User signals: "总结一下", "生成报告", "输出学习记录"
• See "Learning Report Generation" section below

Learning Report Generation

When user requests learning summary, generate TWO files in the appropriate subdirectories and update the index:

File 1: Learning process (learning-process/[date]-[topic]-learning-process.md)

Document the complete learning process:

# [知识点] 学习历程

**学习时间**: [日期]
**学习者**: [如果知道用户名]

## 初始理解水平

[用户在第一个问题中的自我解释原文]

## 评估过程

### 问题 1: [问题内容]
**回答**: [用户回答]
**分析**: [显示出的知识水平/误解]

### 问题 2: [问题内容]
...

## 教学内容记录

### 第一轮教学
**讲解内容**: [概括讲了什么]
**用户疑问**: [用户提出的问题列表]

### 第二轮教学
...

## 理解障碍点

列出学习过程中用户遇到困难的地方：

1. **[障碍点1]**: [什么问题让用户卡住了]
   - 原因分析: [为什么这里难理解]
   - 解决方式: [如何帮助用户突破]

2. **[障碍点2]**: ...

## 学习进度

- ✅ 已掌握: [列出已经理解的概念]
- 🔄 部分理解: [列出还需巩固的部分]
- ❌ 未涉及: [列出还没讲到的高级内容]

File 2: Knowledge Report (knowledge/[date]-[topic]-knowledge.md)

Comprehensive knowledge documentation:

# [知识点] 知识报告

**生成时间**: [日期]

## 概念定义

[清晰、完整的概念定义]

## 核心要点

1. **[要点1]**: [解释]
2. **[要点2]**: [解释]
...

## 详细阐述

### [子主题1]
[详细内容]

### [子主题2]
[详细内容]

## 常见误解 ⚠️

基于本次学习过程，标注容易产生误解的地方：

1. **误解**: [错误理解]
   - **正确理解**: [正确的理解]
   - **为什么容易错**: [解释]

## 进阶方向

如果要继续深入学习，可以探索：
- [进阶主题1]
- [进阶主题2]

## 参考资源

[如果适用，推荐学习资源]

File 3: Index Update (index.md)

After generating the learning report, update the index file with a new entry:

# Knowledge Learning Index

## 学习记录索引

| 日期 | 主题 | 知识报告 | 学习过程 | 关键要点 |
|------|------|----------|----------|----------|
| [YYYY-MM-DD] | [主题] | [链接到知识报告] | [链接到学习过程] | [3-5个关键要点] |

Index Entry Format:

• 日期: 学习日期
• 主题: 学习的知识点
• 知识报告: 相对链接到 knowledge/[date]-[topic]-knowledge.md
• 学习过程: 相对链接到 learning-process/[date]-[topic]-learning-process.md
• 关键要点: 本次学习的核心收获（3-5个要点）

Example Entry:

| 2024-01-15 | 机器学习 | [机器学习知识报告](knowledge/2024-01-15-machine-learning-knowledge.md) | [机器学习学习过程](learning-process/2024-01-15-machine-learning-learning-process.md) | 监督学习vs无监督学习、训练集/测试集划分、过拟合概念 |

Implementation Notes

State Tracking

Maintain awareness of current phase:

• Assessment phase: Counting questions asked (target 5-8)
• Teaching phase: Tracking what's been covered vs. held back
• Interactive phase: Noting recurring confusion points
• Report phase: Compiling complete process

Adaptation Signals

Watch for these user signals:

Signal	Meaning	Response
"不太明白"	Explanation too complex	Simplify, use analogy
"这个我知道"	User has this knowledge	Skip ahead, less detail
"能举个例子吗"	Needs concrete illustration	Provide specific example
"和X有什么关系"	Connecting to prior knowledge	Explain relationship
"继续"	Ready for more	Proceed to next concept
"总结一下"	Wants to conclude	Generate learning report

Quality Checks

Before moving between phases:

• ✅ Assessment: Have I asked 5-8 questions including self-explanation?
• ✅ Teaching: Did I start with fundamentals only?
• ✅ Teaching: Did I explicitly state what I'm NOT covering yet?
• ✅ Interactive: Am I waiting for user questions, not pushing content?
• ✅ Report: Did I capture all confusion points and misconceptions?

Examples

Example 1: Learning "Machine Learning"

Phase 1 - Assessment:

AI: 在我开始教学之前，请你先用自己的话解释一下：你目前对"机器学习"的理解是什么？

User: 我知道机器学习就是让计算机自己学习，不用人编程。好像和AI有关系。

AI: [分析：用户有基本概念，但可能混淆"不用编程"的含义]
你提到"不用人编程"，那你觉得机器学习中，人类需要做什么工作呢？

User: 嗯...可能需要给数据？

AI: 对，数据很重要。那你听说过"训练模型"这个说法吗？知道是什么意思吗？
...

Phase 2 - Teaching:

AI: 好的，基于我们的讨论，我先给你讲解机器学习的核心概念。

机器学习确实是让计算机"学习"，但这里的"不用编程"需要澄清：
我们不是不写代码，而是不需要为每个具体情况写规则...

[3-4段基础解释]

现在我们先理解这个基础部分。关于深度学习、神经网络这些高级话题，我们稍后再讨论。

这部分清楚吗？

Phase 3 - Interactive:

User: 那训练数据是不是越多越好？

AI: 好问题！一般来说数据越多，模型能学到的模式越丰富。但有几个需要注意的点...

Example 2: Learning "Quantum Entanglement"

Phase 1 - Assessment:

AI: 请你先用自己的话说说，你目前对"量子纠缠"的理解是什么？

User: 我只知道好像两个粒子会有神秘的联系，改变一个另一个也会变。

AI: [分析：有基本印象，可能有"超光速通信"误解]
你说的"改变一个另一个也会变"，你觉得这个变化是瞬间发生的吗？

User: 应该是吧？我听说比光速还快。

AI: [确认了误解] 那你觉得这是否意味着我们可以用量子纠缠来实现超光速通信？
...

Anti-Patterns (避免的做法)

❌ Don't dump everything at once: Resist urge to give complete explanation in Phase 2 ❌ Don't skip assessment: Even if topic seems basic, user's mental model may surprise you ❌ Don't ask exam-style questions: "What is X?" feels like testing, not teaching ❌ Don't proceed without user signal: Wait for questions or explicit "continue" before advancing ❌ Don't ignore misconceptions: Address wrong understanding immediately, gently ❌ Don't use jargon prematurely: Introduce technical terms only after explaining concept

Success Criteria

A successful learning session:

• ✅ User can explain concept in their own words
• ✅ User asks questions showing genuine curiosity
• ✅ User corrects their own initial misconceptions
• ✅ User makes connections to their existing knowledge
• ✅ User knows what they don't know yet (明确自己的知识边界)