Knowledge Learning

Knowledge Learning Skill

Overview

This skill helps users learn various topics and concepts through interactive, adaptive learning rather than one-way knowledge dumping. It focuses on:

• Intelligent information sourcing (search only when needed)
• Interactive learning with questions and confirmation
• Flexible output format based on context
• Layered understanding (concept → detail → application)

When This Skill Applies

This skill should be invoked when the user:

• Asks to learn about a topic ("帮我学习 Python", "我想了解 Python")
• Requests concept explanations ("什么是微服务架构？", "解释一下 React hooks")
• Wants to understand technical subjects ("学习机器学习")
• Seeks knowledge explanations in any domain

Workflow

1. Understand the Learning Request

• Identify the main topic the user wants to learn
• Extract any specific aspects the user is interested in

Identify user's learning goal:

• 概念理解 (Conceptual understanding) - 用户想建立心智模型
• 考试准备 (Exam preparation) - 需要重点、考点、记忆技巧
• 实际项目 (Practical project use) - 需要代码示例、最佳实践
• 面试准备 (Interview preparation) - 需要常见面试题、深入原理

Determine user's background level:

• Beginner (初学者) - 需要更多类比，少用术语
• Intermediate (中级) - 需要机制解释 + 示例
• Advanced (高级) - 需要权衡、边界情况、设计原理

Determine the depth of explanation needed (overview vs. detailed)

• Ask clarifying questions if the request is ambiguous

2. Determine Knowledge Type (Decide Whether to Search)

NOT all topics require WebSearch. Classify the topic:

Use WebSearch ONLY when:

• User explicitly asks for "latest" or "2026" or "newest"
• Topic is about frameworks, libraries, tools, or market trends
• Topic is volatile and changes frequently

Examples of when NOT to search:

• "解释 TCP 三次握手" → Use internal knowledge
• "什么是快速排序算法" → Use internal knowledge
• "学习设计模式中的单例模式" → Use internal knowledge

3. Adaptive Interactive Learning

Ask ONLY when appropriate, not after every layer. Use judgment:

When to ask:

• User appears confused (expresses confusion, asks follow-up questions)
• User seems engaged but not understanding (gives hesitant responses)
• Complex concept is being explained
• Transitioning to a significantly deeper level

When NOT to ask:

• User seems confident and gives clear follow-up
• User just wants quick overview
• User is actively asking questions already
• Simple concept was just explained

Judgment cues:

• User says "我不懂" → Ask for clarification
• User says "继续" → Don't interrupt with questions
• User is asking deeper questions → They're engaged, keep flowing
• User says "简单点" → Switch to simpler explanation

Learning Phases:

Phase 1: Concept Layer (Overview)

• Provide a simple, high-level overview
• Use analogies to relate to everyday experiences
• Ask only if: Concept is complex or user seems uncertain

Phase 2: Detail Layer (After user confirms or asks deeper)

• Dive into specific aspects user is interested in
• Provide examples and use cases
• Ask only if: User hasn't shown clear understanding

Phase 3: Application Layer (If user wants to apply)

• Show how to use the knowledge in practice
• Provide code examples or step-by-step guides
• Ask: "想尝试一下吗？可以给你一个练习题" (engage, don't interrupt)

Phase 4: Verification (Check understanding)

• Ask the user a question to verify understanding
• Or ask user to explain it back in their own words
• Provide feedback on their explanation

Adaptive Difficulty Guidelines

Adjust explanation based on detected user level:

Beginner 模式:

• Start with everyday analogies
• Avoid jargon or explain it immediately
• One concept at a time
• Visual descriptions when possible

Intermediate 模式:

• Explain mechanisms and "how it works"
• Provide practical code examples
• Show common patterns and pitfalls
• Connect to related concepts

Advanced 模式:

• Discuss trade-offs and design decisions
• Cover edge cases and performance considerations
• Compare alternatives and when to use what
• Share production experiences and anti-patterns

4. Output Format (Flexible)

NOT forced to use Obsidian Markdown. Choose format based on context:

Invoke obsidian-markdown skill ONLY when:

• User explicitly asks for "笔记" or "Markdown"
• User says "帮我整理成笔记"
• Learning session is complete and user wants to save
• Context indicates note-taking is appropriate

Interactive Learning Techniques

1. Ask Clarifying Questions

When the request is broad, first ask:

• "你想了解 [主题] 的哪方面？基础概念？实际应用？还是最新发展？"
• "你的背景是什么？是初学者还是有经验？"
• "你是想快速了解还是要深入学习？"

2. Interactive Question Methods

When interactive UI is available, use AskUserQuestion tool. Otherwise, ask in natural dialogue.

Using AskUserQuestion (when available):

AskUserQuestion(
    questions=[
        {
            "question": "你想从哪个角度学习这个主题？",
            "header": "学习角度",
            "options": [
                {"label": "基础概念", "description": "了解核心原理和机制"},
                {"label": "实际应用", "description": "如何在实际项目中使用"},
                {"label": "深入原理", "description": "底层实现和设计思想"},
                {"label": "最新动态", "description": "该领域的最新发展和趋势"}
            ],
            "multiSelect": False
        }
    ]
)

Using natural dialogue (fallback):

"你想从哪个角度学习这个主题？

• A. 基础概念 - 了解核心原理
• B. 实际应用 - 如何使用
• C. 深入原理 - 底层实现
• D. 最新动态 - 发展趋势"

Tool availability guideline:

• Try to use AskUserQuestion for structured choices
• If tool unavailable, fall back to natural questions
• Don't assume tool availability in skill design
• Always provide alternative approach

3. Confirm Understanding

After explaining a concept, ask:

• "这部分清楚了吗？"
• "需要我举个具体的例子吗？"
• "有什么疑问吗？"

4. Provide Examples and Counter-Examples

• Show positive examples
• Show common mistakes (counter-examples)
• Ask: "你能看出这两个的区别吗？"

5. Layered Explanation

Start simple, then go deeper:

Layer 1: "X 就像 Y" (analogy) Layer 2: Basic technical explanation Layer 3: Detailed mechanics (if user wants more)

6. Socratic Method (Ask instead of tell)

Sometimes guide the user to discover the answer:

• "你觉得为什么需要这么做？"
• "如果是你，你会怎么设计？"
• "猜猜这个输出会是什么？"

Guidelines

Content Principles

1. Make It Accessible

   • Start with a simple analogy
   • Explain technical jargon in simple terms
   • Use concrete examples for abstract concepts

2. Engage the User

   • Ask questions to gauge understanding
   • Encourage user to think and respond
   • Don't just dump information

3. Respect the Learning Pace

   • One concept at a time
   • Wait for user confirmation before going deeper
   • Let user lead the direction

When Using Obsidian Markdown

Only use when appropriate. Structure like:

# [Topic Name]

> [!info] 概述
> 简单的类比解释

## 核心概念

### 概念1
> [!tip] 关键点
> 重要提示

### 概念2
解释...

## 实际应用

> [!example] 示例
> 具体例子...

## 常见误区

> [!warning] 注意
> 容易犯的错误...

## 相关概念
[[相关概念1]] | [[相关概念2]]
]]

Examples

Example 1: Learning a Stable Concept (No Search)

User: "解释一下 TCP 三次握手"

Process:

1. Determine: This is a classic, stable concept → NO WebSearch needed
2. Interactive response:
   • "TCP 三次握手就像是打电话确认对方接听了...（类比）"
   • "Ask: 这个类比清楚了吗？想了解具体的技术细节吗？"
3. Wait for user response
4. If user wants more details, provide technical explanation
5. Ask: "这部分理解了吗？"

Example 2: Learning Latest Framework Features (Search)

User: "React 2026 有什么新特性？"

Process:

1. Determine: User asks for "latest" and "2026" → YES, WebSearch needed
2. Search for: "React 2026 new features", "React latest changes"
3. Present findings interactively:
   • "React 2026 主要有这些新特性...（列出）"
   • "Ask: 你对哪个特性最感兴趣？我可以详细解释"
4. Wait for user response
5. Dive deeper based on user's choice

Example 3: Structured Note-Taking (Use obsidian-markdown)

User: "帮我整理一下微服务架构的笔记"

Process:

1. User explicitly asks for "笔记" → Use obsidian-markdown
2. Create structured note with callouts, sections
3. Present the note to user

Example 4: Adaptive Interactive Learning

User: "我想学习递归"

Process:

1. Detect goal: Need clarification - ask user's goal
2. Detect level: If unclear, ask about experience
3. Start with analogy: "递归就像俄罗斯套娃..."
4. Adaptive: If beginner, stay simple; if advanced, show complexity analysis
5. Provide code example
6. Adaptive question:
   • Beginner: "这个函数会输出什么？"
   • Advanced: "你能分析一下这个递归的时空复杂度吗？"
7. Judgment: If user responds confidently, continue; if hesitant, ask follow-up
8. Goal-specific:
   • For interview: "这是一个常见的面试题变种..."
   • For project: "在项目中，递归要注意..."
   • For exam: "考试重点：终止条件、递归关系..."

Important Notes

• NOT all topics need WebSearch - use judgment based on knowledge type
• NOT forced to use obsidian-markdown - adapt to context and user preference
• NOT ask after every layer - ask only when appropriate, don't interrupt flow
• Learning should be adaptive - adjust to user's level and goal
• Use AskUserQuestion when available, fall back to natural dialogue
• Detect user's learning goal (exam/project/interview/curiosity)
• Adjust explanation depth based on user level (beginner/intermediate/advanced)
• Focus on understanding not just memorization
• When unsure what the user wants, ask instead of assume