Lecture Notes Master

Lecture Notes Master

Generate structured Obsidian lecture notes with recursive atomic decomposition:

• 主笔记 (Main Note): Hub note with overview, core sections, summary, review questions
• 原子笔记 (Atomic Notes): Deep concept notes in ≥3 layers, each layer with rich structure
• 原子概念 (Glossary): Unlimited bilingual term definitions

When to Apply

Triggers:

• User provides: URL, video link, PDF, transcript, PPT, slides, article
• User says: "总结", "summarize", "create notes", "lecture notes", "笔记", "做笔记"
• User mentions: Obsidian, atomic notes, wikilinks, PARA, MOC
• User provides content for note-taking or analysis
• Exam prep / study materials

Example prompts:

• "总结一下这个视频 https://youtu.be/xxx"
• "Create lecture notes for this article"
• "帮我做笔记 https://www.example.com/article"
• "Summarize this into Obsidian notes"

User Profile

Configured in config.json:

• Name: Schaefer (Zonghan Jia)
• University: Heidelberg University (ZITI), Computer Engineering
• Obsidian Vault: See config.json → obsidian.vault_path
• Output Directory: 00-Inbox/{Topic}/ (main note + glossary + numbered L1 subdirectories)
• Language: Bilingual — English primary, Chinese secondary
• Term format: English Term（中文术语）

Core Principles

Recursive Atomic Decomposition

Every source material is decomposed into a tree of notes, organized into numbered subdirectories by L1 topic:

{Topic}/
├── 主笔记: {Topic}-Notes.md
│   Hub note linking to all L1 atomic notes
│
├── 01-{L1-Concept-A}/
│   ├── {L1-Concept-A}.md              (L1 顶层概念)
│   ├── {L2-Sub-Concept-A1}.md         (L2 子概念)
│   ├── {L2-Sub-Concept-A2}.md         (L2 子概念)
│   ├── {L3-Detail-A1a}.md             (L3 细分解)
│   └── {L3-Detail-A1b}.md             (L3 细分解)
│
├── 02-{L1-Concept-B}/
│   ├── {L1-Concept-B}.md              (L1 顶层概念)
│   ├── {L2-Sub-Concept-B1}.md         (L2 子概念)
│   └── {L3-Detail-B1a}.md             (L3 细分解)
│
├── 03-{L1-Concept-C}/
│   └── {L1-Concept-C}.md              (L1 顶层概念)
│
├── glossary/（原子概念 — 术语定义，不限量）
│   ├── English Term（中文术语）.md
│   └── ... 每个术语一个文件
│
└── assets/（图表资源）
    └── *.png

Directory Organization Rules

Numbering Rules:

1. L1 subdirectories use two-digit prefix: 01-, 02-, 03-, ...
2. No atomic notes in the topic root — only the main hub note lives there
3. Within each L1 folder, all notes are flat (L1 + L2 + L3 together, no further nesting)
4. Numbering follows the order of L1 concepts as they appear in the main note
5. Wiki-links use filename only (no path prefix) — Obsidian resolves them automatically

Decomposition Rules

Layer 1 (顶层概念):

• Source material's major themes, sections, or arguments
• Each L1 note covers ONE major concept branch
• Number: determined by content (typically 3-7, NOT fixed)

Layer 2 (子概念):

• Sub-concepts, mechanisms, or components within each L1 concept
• Each L2 explains a specific aspect of its parent L1
• Number: typically 2-4 per L1 parent

Layer 3+ (细分解):

• Specific mechanisms, case studies, comparisons, or evidence
• Finest granularity of analysis
• Number: as many as the content demands

Stop Decomposing When:

• A concept can be fully explained in ≤500 words with one diagram
• Further splitting would break logical coherence
• The concept is better served as a glossary entry (pure definition)

Glossary vs Atomic Note:

Minimum Requirement

Every set of notes MUST produce:

• 1 main note (主笔记)
• ≥3 layers of atomic notes (L1 → L2 → L3 minimum)
• Glossary entries for ALL technical terms mentioned

Every Atomic Note Must Be Rich

ALL atomic notes (L1, L2, L3) use the same rich template structure:

1. 定义 — One-paragraph definition with inline wikilinks
2. Why Do We Need This? — Motivation with concrete scenario
3. Core Concept — Idea + example + Mermaid diagram + step-by-step breakdown
4. Comparison — Table comparing with/without, or vs alternatives
5. Common Pitfalls — 2-3 mistakes and fixes (omit only if truly not applicable)
6. Key Takeaway — One flashcard-worthy sentence
7. Review Questions — 3 levels: recall, understanding, application
8. Related Notes — Parent (UP), children (DOWN), siblings (ACROSS)

See templates/atomic-note.md for the full template.

Writing Style: Runoob Tutorial（菜鸟教程风格）

Prerequisites

python3 --version || python --version
pip3 install matplotlib numpy  # For visualization scripts (optional)

Optional: summarize CLI for URL/video content extraction.

Workflow

Step 1: Analyze User Input

Identify: input type (URL, video, PDF, transcript, raw text), topic name, key themes.

URL & Video Content Extraction

When user provides a URL or video link, use summarize CLI:

# Extract text from URL
summarize "<URL>" --extract-only --model google/gemini-3-flash-preview

# Extract YouTube transcript
summarize "<YouTube-URL>" --youtube auto --extract-only

# Pre-summary for screening
summarize "<URL>" --length medium --model google/gemini-3-flash-preview

Useful flags: --extract-only (raw text), --youtube auto, --firecrawl auto (JS-heavy sites), --json

If summarize is not available: Use the agent's built-in web fetching tools as fallback.

Step 2: Plan Decomposition Tree

⚠️ MANDATORY: You MUST complete this step BEFORE writing ANY content. ⚠️ MANDATORY: You MUST create ALL directories with mkdir -p BEFORE writing ANY files. ⚠️ ZERO TOLERANCE: No atomic notes are allowed in the topic root directory. ONLY the main hub note lives there.

Plan the COMPLETE file tree with numbered directories, then create them immediately.

2a. Plan the tree (output to user for confirmation)

Topic: 2028 Global Intelligence Crisis

主笔记: 2028-Global-Intelligence-Crisis-Notes.md  (in topic root)

01-Intelligence-Displacement-Spiral/
  L1: Intelligence-Displacement-Spiral.md
  L2: OpEx-Substitution-Mechanism.md
  L2: Ghost-GDP-Phenomenon.md
  L3: OpEx-vs-CapEx-AI-Spending.md               ← parent: OpEx-Substitution
  L3: Why-No-Natural-Brake.md                     ← parent: OpEx-Substitution

02-SaaS-Collapse-and-Intermediation-Death/
  L1: SaaS-Collapse-and-Intermediation-Death.md
  L2: Agentic-Coding-Disruption.md
  L2: Habitual-Intermediation-Collapse.md
  L3: Friction-Zero-Disruption.md                 ← parent: Habitual-Intermediation

03-White-Collar-Displacement-Asymmetry/
  L1: White-Collar-Displacement-Asymmetry.md
  L2: Downshifting-Effect.md
  L2: Labor-Share-Decline.md

04-Financial-Contagion-Chain/
  L1: Financial-Contagion-Chain.md
  L2: Private-Credit-SaaS-Crisis.md
  L2: Permanent-Capital-Trap.md
  L2: Mortgage-Market-Structural-Threat.md
  L3: Zendesk-Case-Study.md                       ← parent: Private-Credit
  L3: Insurance-Asset-Impairment.md               ← parent: Permanent-Capital

glossary/ (42 entries):
  glossary/Ghost GDP（幽灵GDP）.md
  glossary/Intelligence Displacement Spiral（智能替代螺旋）.md
  glossary/Private Credit（私募信贷）.md
  ... (one file per term)

2b. Create ALL directories FIRST (before writing any files)

# MANDATORY: Run this BEFORE writing any notes
TOPIC_DIR="<vault>/00-Inbox/{Topic}"
mkdir -p "$TOPIC_DIR"
mkdir -p "$TOPIC_DIR/01-{L1-Concept-A}"
mkdir -p "$TOPIC_DIR/02-{L1-Concept-B}"
mkdir -p "$TOPIC_DIR/03-{L1-Concept-C}"
# ... one mkdir per L1 concept
mkdir -p "$TOPIC_DIR/glossary"
mkdir -p "$TOPIC_DIR/assets"

2c. Verify directory structure before proceeding

# Verify: must show numbered subdirectories + glossary + assets
find "$TOPIC_DIR" -type d | sort

Only proceed to Step 3 after directories exist.

Step 3: Search Existing Resources

# Search glossary for existing bilingual terms
python3 <SKILL_DIR>/scripts/search.py "<topic keywords>" --domain glossary

# Search Mermaid templates for diagram ideas
python3 <SKILL_DIR>/scripts/search.py "<concept type>" --domain mermaid

# Search writing rules for style guidance
python3 <SKILL_DIR>/scripts/search.py "<content type>" --domain writing

Step 4: Generate Notes (in order)

⚠️ CRITICAL: Every file MUST be written to its correct subdirectory. Double-check the output path before every Write/Edit call.

Self-check before each file write:

• Is this the main hub note? → Write to {Topic}/ root
• Is this an atomic note (L1/L2/L3)? → Write to {Topic}/NN-{L1-Parent}/
• Is this a glossary entry? → Write to {Topic}/glossary/
• NEVER write atomic notes directly to {Topic}/ root

Script Usage

# Main note (in topic root)
python3 <SKILL_DIR>/scripts/generate.py \
  --type lecture --title "<title>" \
  --concepts "C1,C2,C3,C4" \
  --output "<vault>/00-Inbox/{Topic}/"

# Atomic note (any layer — output to its L1 parent subdirectory)
python3 <SKILL_DIR>/scripts/generate.py \
  --type atomic --concept "<name>" \
  --parent "<parent-note-stem>" \
  --children "Child1,Child2" \
  --siblings "Sibling1,Sibling2" \
  --output "<vault>/00-Inbox/{Topic}/NN-{L1-Concept}/"

# Glossary entry
python3 <SKILL_DIR>/scripts/generate.py \
  --type glossary --term-en "<English Term>" --term-cn "<中文术语>" \
  --definition "<one-line definition>" \
  --output "<vault>/00-Inbox/{Topic}/glossary/"

# Course MOC
python3 <SKILL_DIR>/scripts/generate.py \
  --type moc --course "<course>" --semester "<semester>"

Note: <SKILL_DIR> = the directory where this skill is installed. NN = two-digit L1 index (01, 02, ...).

Step 5: Generate Visualizations (as needed)

For data-driven charts that Mermaid cannot handle:

python3 <SKILL_DIR>/scripts/visualize.py \
  --type "<chart_type>" \
  --data "<data_json>" \
  --output "<output_path>" \
  --title "<chart title>"

Available chart types: bar, line, scatter, timeline, heatmap, comparison, pie, radar

Output Format Templates

All templates are in the templates/ directory:

Main Note (主笔记) Key Requirements

• YAML frontmatter: title, date, course/topic, tags, source_files, status, aliases
• Core Idea（核心思想）blockquote with inline [[glossary-wikilinks]]
• Source links
• Overview section (with image embed if available)
• N numbered sections (one per L1 concept, NOT limited to 3), each with:
  • → 详见 [[L1-Atomic-Note]] link
  • Sub-sections: Why → What → How
  • Mermaid diagrams, comparison tables, key insight blockquotes
• Summary section: concept map (Mermaid) + data summary table + key quote
• Review Questions: recall + understanding + application + critical thinking
• Related Notes linking to all L1 atomic notes