course-reading-notes

name	course-reading-notes
description	Multi-course academic reading notes generator — paper selection, note writing, Nature-style figures, Chinese PDF, and submission packaging
keywords	["课程作业","阅读笔记","论文阅读","course-assignment","reading-notes","academic-report","课程报告","论文笔记"]

Course Reading Notes Skill

Generate complete, publication-quality course reading notes from scratch: paper selection, note writing, figures, Chinese PDF, and submission packaging. Multi-course generic.

When to use

• User asks for a course reading assignment, paper reading notes, or course report
• Keywords: "课程作业", "阅读笔记", "论文阅读", "course assignment", "reading report"
• Any course requiring structured academic notes with figures and PDF output

When NOT to use

• Only PDF generation needed (use chinese-pdf skill instead)
• Only figure generation needed (use nature-figure skill instead)
• Full paper reading/translation (use nature-reader skill instead)

Workflow

Phase 1: Paper Selection

If user hasn't selected a paper, help find one. Criteria:

1. Recency: Published within the last 5 years
2. Venue: Top conferences (AAAI, IJCAI, NeurIPS, ICML, GECCO, PPSN, CEC) or top journals
3. Open access: PDF freely downloadable (arxiv preferred)
4. Open source code: GitHub repo with actual implementation, not placeholder
5. Course relevance: Matches the course topic

Verification steps:

• Check GitHub repo has real code (stars, file structure, actual .py files)
• Verify arxiv PDF downloads successfully
• Extract paper title, authors, venue, year, abstract

Phase 2: Write Reading Notes

Structure (adaptable to course requirements):

# 论文阅读笔记
<center><b>学号 + 姓名</b></center>

## 论文基本信息
- **Title**: 论文英文标题
- **作者**: 作者列表
- **单位**: 作者单位
- **会议/期刊**: 发表 venue + year
- **关键词**: 3-5个关键词

## 一、摘要
论文核心问题、方法、结果的精炼总结（200-300字）

## 二、背景与动机
### 2.1 问题背景
### 2.2 现有方法局限
### 2.3 本文思路

## 三、方法
### 3.1 [核心方法1]
### 3.2 [核心方法2]
### 3.3 [训练/优化机制]

## 四、实验
### 4.1 实验设置
### 4.2 主要结果
### 4.3 [其他实验]

## 五、分析与思考
### 5.1 与课程内容的联系
### 5.2 优点
### 5.3 局限性

## 六、总结
2-3段总结性文字

## 参考资料
- 论文引用信息
- 开源代码链接

Phase 3: Generate Figures

Create 3-4 Nature-style figures for the notes. Common figure types:

Figure Type	Purpose
Overview diagram	Conceptual comparison or method overview
Architecture diagram	Method/algorithm structure
Hierarchical tree	Modular structure or taxonomy
Line plots	Performance comparison curves
Bar charts	Quantitative results comparison

Mandatory font loading (every figure script must start with):

import matplotlib.font_manager as fm
fm._load_fontmanager(try_read_cache=False)
CJK_FONT = '/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc'
prop = fm.FontProperties(fname=CJK_FONT, size=14)

Then pass fontproperties=prop to every text element (title, label, annotation, legend):

ax.set_title('标题', fontproperties=prop)
ax.set_xlabel('X轴', fontproperties=prop)
ax.text(x, y, '文本', fontproperties=prop)
ax.legend(prop=prop)

Never rely on plt.rcParams['font.sans-serif'] alone — it silently falls back to fonts without CJK glyphs, producing `` tofu boxes.

Other mandatory rules:

• Replace Unicode subscript/superscript with plain ASCII (S1 not S₁)
• Max image width 1000px for PDF embedding
• Image markdown: ![](path/to/image.png){width=85%} — empty alt text to avoid duplicate captions
• Figure caption as italic text below image: *图X：图注文字*
• Clean layout, restrained color palette (Nature style)

Common pitfalls:

• Never use ![图X：...](path) — pandoc generates duplicate "Figure X: 图X：" labels
• Always use ![](path){width=85%} with empty alt text
• Image files must be in 相关资料/ subdirectory
• Never use family='monospace' for Chinese text (monospace fonts lack CJK glyphs)

Phase 4: Compile PDF

Markdown YAML frontmatter:

---
title: 论文阅读笔记
author: 学号 + 姓名
date: 2026年X月
documentclass: ctexart
geometry: margin=2.8cm
fontsize: 11pt
colorlinks: true
header-includes: |
  \setlength{\emergencystretch}{3em}
---

Compilation command:

cd <作业目录>
pandoc 论文阅读笔记.md -o 论文阅读笔记.pdf --pdf-engine=xelatex --shift-heading-level-by=0

Critical:

• Must use ctexart, NOT article + manual mainfont
• margin=2.8cm prevents text hugging page edge
• emergencystretch=3em helps Chinese long-sentence line-breaking

Phase 5: Package Submission

# Clean old figures that don't belong to current assignment
rm -f 相关资料/fig*_old.png

# Create zip
rm -f 作业名.zip
zip -r 作业名.zip 论文阅读笔记.md 论文阅读笔记.pdf 相关资料/

Verification Checklist

Before declaring complete:

• PDF has no right-side text overflow (check x1 < page_width)
• No duplicate "Figure X:" labels in PDF
• All images render correctly with Chinese text
• Markdown alt text is empty for all images
• Each image has {width=85%} attribute
• Submission zip contains only current assignment files

Example Usage

User: "这是我的课程作业要求，帮我完成一份论文阅读笔记"
Assistant: [invokes course-reading-notes skill]
  1. Find appropriate paper
  2. Write structured notes
  3. Generate 4 figures
  4. Compile PDF
  5. Package submission