Lecture notes enhancement

Lecture Notes Enhancement

Shared narrative rules

Read and follow the shared narrative and evidence rules. Keep process notes and defensive source disclaimers out of the lecture notes; state source limitations only when they affect the explained concept.

Overview

Systematic process for cross-referencing recordings, PPTs, and existing notes to identify gaps and enhance lecture notes with narrative-first writing. Uses parallel subagents for efficiency.

When to Use

Use when:

• Enhancing existing lecture notes (讲义.md) with course recordings and PPTs
• Course has multiple source materials (records/, extracted/, existing notes)
• Need systematic gap identification and content supplementation
• Following md-report-writer narrative style (narrative flow before technical content)

Don't use when:

• Creating lecture notes from scratch (use brainstorming + writing-plans instead)
• Simple formatting or proofreading (use simplify skill instead)
• Single source material only (direct editing is sufficient)

Nature of Source Materials

Each source type has distinct role and limitations:

Slides (PPT extracted/) — Simplified outline only. Contain key formulas and section outlines — the final results, not the path to them. No detailed derivations, intermediate steps, or physical intuition. Treat slides as a what checklist, not the full content.

Recordings (records/) — Full content with caveats. Contain detailed derivations, step-by-step reasoning, and additional formulas not on slides. However: (a) teachers may not read every formula aloud — extract keywords to identify which formulas are being discussed; (b) ASR transcriptions introduce errors, especially for technical terms — always cross-reference with slides for precise formulas.

Existing notes — Baseline. Provide current structure and coverage level to supplement, not replace.

Core Pattern

The enhancement process follows four phases:

Phase 1: Parallel Analysis - Dispatch subagents to analyze each source material independently Phase 2: Gap Synthesis - Identify missing content, inconsistencies, and improvement opportunities Phase 3: Systematic Enhancement - Edit lecture notes section-by-section with narrative-first style Phase 4: Verification - Ensure all improvements follow md-report-writer guidelines

Phase 1: Parallel Analysis Dispatch

For each course record or PPT file, dispatch a subagent with specific analysis scope:

Analyze records/week03.md:
- Focus: Numerical methods for PDE solving
- Extract: Discretization schemes, boundary conditions, solver algorithms
- Output: Structured summary with key formulas and implementation details

Why parallel? Each source is independent. Processing concurrently saves time and prevents context overflow.

Phase 2: Gap Identification Matrix

Create structured gap analysis organized by priority:

Priority	Content Category	Missing Details	Source Location
High	Theoretical foundations	Fundamental equation derivation	records/week02.md L30-40
High	Numerical methods	Iterative solver details	records/week03.md L170-190
Medium	Implementation tips	Optimization techniques	records/week04.md L330-360

Phase 3: Narrative-First Enhancement

Apply md-report-writer style principles:

✅ DO:

• Start sections with explanatory narrative
• Use bold for emphasis (never italics)
• Write complete, grammatical sentences
• Embed code snippets to support claims
• Weave narrative through content (narrative → content → narrative)

❌ DON'T:

• Start sections abruptly with lists/formulas
• Use bullet points for main narrative explanations
• Use conversational filler ("we will explore", "as we can see")
• Use vague emphasis words ("核心", "关键", "重要", "主要", "本质", "显著") as a substitute for evidence — state the role, input/output, or measured effect instead
• Write multi-paragraph docstrings or comment blocks
• Add meta-commentary about the editing process, source fidelity, or why a sentence was written a certain way
• Explain that an unsupported detail cannot be inferred unless that limitation changes the lecture content

Segment Length Principle

Break complex explanations into focused segments, giving the user a chance to respond after each. If answering 3+ distinct questions, deliver them as separate segments. Avoid massive single responses.

Phase 4: Section-by-Section Editing

Edit the lecture notes systematically by section:

1. Read current section in existing lecture notes
2. Read relevant analysis from supplement reports
3. Identify specific improvement needed
4. Apply narrative-first edit using Edit tool
5. Update todo list to track progress

Quick Reference

Source Type	What to Extract	Typical Locations	Limitations
Course records (records/)	Physical intuition, detailed derivations, additional formulas not on slides	Usually lines 50-200	Transcription inaccuracies; teachers may not read every formula verbatim
PPT content (extracted/)	Key formulas, section structure, topic checklist	Throughout file	Simplified only — missing derivations, intuition, and content spoken in class
Existing notes	Current structure, what needs enhancement	Full file	May inherit gaps from whichever source was used previously

Common Enhancement Categories

Theoretical Foundations: Missing derivations, physical intuition, intermediate steps between slide formulas.

Numerical Methods: Solver details, algorithm choices, convergence analysis not shown on slides.

Implementation Details: Practical considerations, edge cases, boundary conditions discussed orally.

Writing Style: Narrative-first structure, symbol definitions at first use, bold for key terms, complete sentences.

Implementation

## Example Enhancement: Adding Cauchy Momentum Equation

**Before** (abrupt start):
### 1.1 Navier-Stokes方程
流体运动的数学描述建立在连续介质力学假设之上。Navier-Stokes方程是...

**After** (narrative-first):
### 1.1 连续介质力学基础
流体仿真的理论基础建立在**连续介质假设**之上：将流体视为连续分布的物质，而非离散的分子集合。在这一假设下，我们关注流体中每个**小体积元**的受力与运动。对于连续介质中的小体积元，它受到的力可以分为两大类。**体力**（body force）作用于整个体积，如重力场和磁场。**表面力**（surface force）通过表面接触传递，如流体内部的摩擦和压力。柯西动量方程（Cauchy Momentum Equation）是适用于所有连续介质的普适方程...

Common Mistakes

Mistake	Why Bad	Fix
Editing without reading current content	May duplicate or miss context	Always Read before Edit
Answering questions without editing notes	User needs enhanced documentation, not conversation	Edit the lecture notes directly
Adding too much content at once	Hard to verify, high error risk	Edit section-by-section with todo tracking
Forgetting narrative flow	Technical content feels abrupt	Start each section with explanatory text
Using italics for emphasis	Md-report-writer specifies bold only	Replace all italics with bold
Writing conversational filler	Unprofessional, wastes tokens	Use direct, professional tone
Skipping parallel analysis	Sequential processing is slower, loses context	Dispatch subagents for independent sources
Leaving symbols unexplained	Readers can't follow formulas	Always define $w_i$, $w_p$, and other notation at first use
Treating slides as complete content	Misses majority of lecture material	Slides provide what topics, recordings provide how they're explained
Trusting transcription text blindly	ASR errors in technical terms	Cross-reference with slides; use recording for explanation, slides for precise formulas
Expecting formulas to be spoken verbatim in recordings	Teachers gesture at slides, don't read every symbol	Extract keywords to identify which formulas are being discussed

Real-World Impact

Before: Lecture notes missing key theoretical foundations, numerical details, and implementation context that only existed in recordings.

After: Enhanced notes with full derivations, physical intuition, and practical implementation details — all following narrative-first style. Each section now explains why before what.

Red Flags - STOP and Re-Read

• About to Edit without Reading current section first
• About to answer a question without editing the notes
• Starting a section with a list or formula (no narrative intro)
• Using italics instead of bold for emphasis
• Writing "we will explore" or "as we can see"
• Processing course records sequentially instead of in parallel
• Editing entire document at once without todo tracking
• About to use a symbol ($w_i$, $\mathbf{u}_i$, etc.) without defining it first
• Assuming slide content is the complete lecture — slides are outlines, not full content
• Copying transcription text into notes without cross-referencing slides for accuracy
• Expecting every formula to be read aloud in the recording — extract keywords instead

All of these mean: Stop, re-read the skill, and follow the correct approach.