playbooks

home / skills / bahayonghang / my-claude-code-settings / latex-thesis-zh

latex-thesis-zh skill

/content/skills/academic-skills/latex-thesis-zh

This skill helps you refine Chinese LaTeX thesis writing, ensuring structure, style, and formatting while preserving citations and math.

npx playbooks add skill bahayonghang/my-claude-code-settings --skill latex-thesis-zh

Review the files below or copy the command above to add this skill to your agents.

Files (25)
SKILL.md
9.6 KB
---
name: latex-thesis-zh
category: academic-writing
tags:
  - latex
  - thesis
  - chinese
  - phd
  - master
  - xelatex
  - gb7714
  - thuthesis
  - pkuthss
  - compilation
  - bibliography
description: |
  中文学位论文 LaTeX 助手（博士/硕士论文）。
  Use when writing, reviewing, or improving Chinese LaTeX theses (PhD/Master).
  Use when user mentions 编译, 格式, 去AI化, 标题, 参考文献, 模板, 一致性,
  or any Chinese thesis quality task.
  支持多种学校模板和 GB/T 7714 国标格式。

  触发词（可独立调用任意模块）：
  - "compile", "编译", "xelatex" → 编译模块
  - "structure", "结构", "映射" → 结构映射模块
  - "format", "格式", "国标", "GB/T" → 国标格式检查模块
  - "expression", "表达", "润色", "学术表达" → 学术表达模块
  - "logic", "coherence", "逻辑", "衔接", "methodology", "方法论" → 逻辑衔接与方法论深度模块
  - "long sentence", "长句", "拆解" → 长难句分析模块
  - "bib", "bibliography", "参考文献" → 参考文献模块
  - "template", "模板", "thuthesis", "pkuthss" → 模板检测模块
  - "deai", "去AI化", "人性化", "降低AI痕迹" → 去AI化编辑模块
  - "title", "标题", "标题优化", "生成标题" → 标题优化模块
  - "consistency", "一致性", "术语" → 一致性检查模块
argument-hint: "[main.tex] [--section <章节>] [--module <模块>]"
allowed-tools: Read, Glob, Grep, Bash(python *), Bash(xelatex *), Bash(lualatex *), Bash(latexmk *), Bash(bibtex *), Bash(biber *)
---

# LaTeX 中文学位论文助手

## 核心原则

1. 绝不修改 `\cite{}`、`\ref{}`、`\label{}`、公式环境内的内容
2. 绝不凭空捏造参考文献条目
3. 绝不在未经许可的情况下修改专业术语
4. 始终先以注释形式输出修改建议
5. 中文文档必须使用 XeLaTeX 或 LuaLaTeX 编译

## 参数约定（$ARGUMENTS）

- `$ARGUMENTS` 用于接收主文件路径、目标章节、模块选择等关键信息。
- 若 `$ARGUMENTS` 缺失或含糊，先询问：主 `.tex` 路径、目标范围、所需模块。
- 路径按字面处理，不推断或补全未提供的路径。

## 执行约束

- 仅在用户明确要求时执行脚本/编译命令。
- 涉及清理（`--clean` / `--clean-all`）等破坏性操作前先确认。

## 统一输出协议（全部模块）

每条建议必须包含固定字段：
- **严重级别**：Critical / Major / Minor
- **优先级**：P0（阻断）/ P1（重要）/ P2（可改进）

```latex
% <模块>（第<N>行）[Severity: <Critical|Major|Minor>] [Priority: <P0|P1|P2>]: <问题概述>
% 原文：...
% 修改后：...
% 理由：...
% ⚠️ 【待补证】：<需要证据/数据时标记>
```

## 失败处理（全局）

```latex
% ERROR [Severity: Critical] [Priority: P0]: <简要错误>
% 原因：<缺少脚本/工具或路径无效>
% 建议：<安装工具/核对路径/重试命令>
```

## 模块（独立调用）

除"结构映射"在**完整审查或多文件场景**中要求先执行外，其余模块均可独立调用。

---

### 模块：编译
**触发词**: compile, 编译, build, xelatex, lualatex

```bash
python scripts/compile.py main.tex                          # 默认: latexmk + XeLaTeX（推荐）
python scripts/compile.py main.tex --recipe xelatex         # XeLaTeX 单次
python scripts/compile.py main.tex --recipe lualatex        # LuaLaTeX 单次
python scripts/compile.py main.tex --recipe xelatex-bibtex  # 传统 BibTeX
python scripts/compile.py main.tex --recipe xelatex-biber   # 现代 biblatex（推荐）
python scripts/compile.py main.tex --outdir build           # 指定输出目录
python scripts/compile.py main.tex --watch                  # 监视模式
python scripts/compile.py main.tex --clean                  # 清理辅助文件
```

自动检测 ctex、xeCJK 或中文字符时选择 XeLaTeX。详见 [COMPILATION.md](resources/COMPILATION.md)

---

### 模块：结构映射
**触发词**: structure, 结构, 映射, map

**完整审查/多文件场景先执行**：分析多文件论文结构。

```bash
python scripts/map_structure.py main.tex
```

输出文件树结构、模板类型检测、章节处理顺序。详见 [STRUCTURE_GUIDE.md](resources/STRUCTURE_GUIDE.md)

---

### 模块：国标格式检查
**触发词**: format, 格式, 国标, GB/T, 7714

```bash
python scripts/check_format.py main.tex
python scripts/check_format.py main.tex --strict
```

检查 GB/T 7714-2015 参考文献格式、图表标题、公式编号、标题样式。详见 [GB_STANDARD.md](resources/GB_STANDARD.md)

---

### 模块：学术表达
**触发词**: expression, 表达, 润色, 学术表达, 口语化

用户提供段落源码，Agent 分析并返回润色版本及对比表格。

**核心规则**：
- 口语 → 学术转换（"很多研究表明" → "大量研究表明"）
- 禁用主观词汇（"显然"、"毫无疑问" → "研究表明"、"实验结果显示"）

详见 [ACADEMIC_STYLE_ZH.md](resources/ACADEMIC_STYLE_ZH.md)

---

### 模块：逻辑衔接与方法论深度
**触发词**: logic, coherence, 逻辑, 衔接, methodology, 方法论, 论证, argument

确保段落间逻辑流畅，强化方法论的严谨性。

**重点检查**：
- 段落级逻辑衔接（AXES 模型：主张→例证→解释→意义）
- 过渡信号词使用
- 方法论深度（方法选择论证、局限性承认、假设陈述）

详见 [LOGIC_COHERENCE.md](resources/LOGIC_COHERENCE.md)

---

### 模块：长难句分析
**触发词**: long sentence, 长句, 拆解, simplify

**触发条件**: 句子 >60 字 或 >3 个从句

```latex
% 长难句检测（第45行，共87字）[Severity: Minor] [Priority: P2]
% 主干：本文方法在多个数据集上取得优异性能。
% 修饰成分：
%   - [定语] 基于深度学习的
%   - [方式] 通过引入注意力机制
% 建议改写：拆分为多个短句，每句聚焦一个信息点。
```

---

### 模块：参考文献
**触发词**: bib, bibliography, 参考文献, citation, 引用

```bash
python scripts/verify_bib.py references.bib
python scripts/verify_bib.py references.bib --tex main.tex    # 检查引用
python scripts/verify_bib.py references.bib --standard gb7714 # 国标检查
```

检查必填字段完整性、重复条目、未使用条目、缺失引用、GB/T 7714 格式合规。

---

### 模块：模板检测
**触发词**: template, 模板, thuthesis, pkuthss, ustcthesis, fduthesis

```bash
python scripts/detect_template.py main.tex
```

| 模板 | 学校 |
|------|------|
| thuthesis | 清华大学 |
| pkuthss | 北京大学 |
| ustcthesis | 中国科学技术大学 |
| fduthesis | 复旦大学 |
| ctexbook | 通用（GB/T 7713.1-2006） |

详见 [UNIVERSITIES/](resources/UNIVERSITIES/)

---

### 模块：去AI化编辑
**触发词**: deai, 去AI化, 人性化, 降低AI痕迹, 自然化

在保持 LaTeX 语法和技术准确性的前提下，降低 AI 写作痕迹。

```bash
python scripts/deai_check.py main.tex --section introduction  # 单章节分析
python scripts/deai_check.py main.tex --analyze               # 全文密度检测
python scripts/deai_check.py main.tex --fix-suggestions        # JSON 修复建议
```

**AI 痕迹检测类型**：空话口号、过度确定、机械排比、模板表达

**硬性约束**：
- **绝不修改**：`\cite{}`、`\ref{}`、`\label{}`、公式环境
- **绝不新增**：事实、数据、结论、指标、实验设置、引用编号
- **仅修改**：普通段落文字、章节标题内的中文表达

详见 [DEAI_GUIDE.md](resources/DEAI_GUIDE.md)

---

### 模块：标题优化
**触发词**: title, 标题, 标题优化, 生成标题, 改进标题

根据学位论文规范和学术最佳实践，生成和优化论文标题。

```bash
python scripts/optimize_title.py main.tex --generate   # 生成候选标题
python scripts/optimize_title.py main.tex --optimize    # 优化现有标题
python scripts/optimize_title.py main.tex --check       # 质量评分（0-100）
```

评分维度：简洁性(25%)、可搜索性(30%)、长度(15%)、具体性(20%)、规范性(10%)

详见 [TITLE_OPTIMIZATION.md](resources/TITLE_OPTIMIZATION.md)

---

### 模块：一致性检查
**触发词**: consistency, 一致性, 术语, terminology

检查全文术语使用一致性。

```bash
python scripts/check_consistency.py main.tex                    # 默认术语组检查
python scripts/check_consistency.py main.tex --custom-terms terms.json  # 自定义术语组
```

**检查项目**：
- 同一概念的中英文术语是否全文统一
- 缩写首次出现是否有全称定义
- 数字/单位格式一致性

---

## 完整工作流（可选）

如需完整审查，按顺序执行：

1. **结构映射** → 分析论文结构
2. **国标格式检查** → 修复格式问题
3. **去AI化编辑** → 降低 AI 写作痕迹
4. **学术表达** → 改进表达
5. **长难句分析** → 简化复杂句
6. **参考文献** → 验证引用

---

## 参考文档

- [STRUCTURE_GUIDE.md](resources/STRUCTURE_GUIDE.md): 论文结构要求
- [GB_STANDARD.md](resources/GB_STANDARD.md): GB/T 7714 格式规范
- [ACADEMIC_STYLE_ZH.md](resources/ACADEMIC_STYLE_ZH.md): 中文学术写作规范
- [FORBIDDEN_TERMS.md](resources/FORBIDDEN_TERMS.md): 受保护术语
- [COMPILATION.md](resources/COMPILATION.md): XeLaTeX/LuaLaTeX 编译指南
- [UNIVERSITIES/](resources/UNIVERSITIES/): 学校模板指南
- [DEAI_GUIDE.md](resources/DEAI_GUIDE.md): 去AI化写作指南
- [TITLE_OPTIMIZATION.md](resources/TITLE_OPTIMIZATION.md): 标题优化详细指南
- [LOGIC_COHERENCE.md](resources/LOGIC_COHERENCE.md): 逻辑衔接与方法论深度指南
- [WRITING_PHILOSOPHY_ZH.md](resources/WRITING_PHILOSOPHY_ZH.md): 学位论文写作哲学

Overview

This skill is a Chinese LaTeX thesis assistant tailored for doctoral and master's theses in deep learning, time series, and industrial control. It provides modular tools for compilation, structure mapping, GB/T format checks, academic polishing, logical coherence, long-sentence analysis, bibliography validation, template detection, de-AIfication, and title optimization. The skill enforces conservative editing rules to preserve citations, labels, and math environments and outputs change suggestions as annotated LaTeX comments. It is designed for iterative, auditable edits and integrates optional scripts for automated checks and compilation when explicitly requested.

How this skill works

You invoke specific modules via trigger words (e.g., "compile", "structure", "format", "expression", "logic", "long sentence", "bib", "template", "deai", "title"). Each module inspects the LaTeX source and returns annotated suggestions in a fixed comment template that includes severity and priority. Compilation routines use latexmk with XeLaTeX/LuaLaTeX heuristics for Chinese documents; analysis modules parse source to avoid touching \cite{}, \ref{}, \label{}, and math environments. Destructive actions (clean, full rebuild) and script execution only run when you explicitly confirm.

When to use it

  • Prepare or proof a Chinese thesis in deep learning, time series, or industrial control.
  • Need safe automated compilation and dependency handling for XeLaTeX/LuaLaTeX workflows.
  • Validate compliance with GB/T bibliographic and formatting conventions.
  • Polish academic phrasing, improve paragraph-level logic, or reduce AI-style wording.
  • Generate or optimize thesis titles to maximize clarity and discoverability.

Best practices

  • Always provide the main .tex path and target chapter when requesting edits; the tool will ask if arguments are missing.
  • Run structure mapping before full-document reviews or when multiple files are involved.
  • Request compilation or destructive clean operations only after reviewing suggested commands and confirming intent.
  • Supply source snippets (preserve indentation) for expression, de-AIfication, or long-sentence modules to get minimal-intrusion edits.
  • Treat all outputs as annotated suggestions; apply changes to source yourself or authorize the script to write files.

Example use cases

  • Compile a Chinese thesis with latexmk + XeLaTeX, auto-running biber/bibtex as needed and returning first-error diagnostics.
  • Map multi-file thesis structure and detect the applied university template and missing front/back matter.
  • Check references.bib for missing required fields, duplicates, and GB/T 7714 compliance while flagging unused or missing cites.
  • Polish a paragraph from colloquial Chinese to formal academic Chinese while preserving citations and math.
  • Analyze a long, 90-character sentence and propose a split with a clear main clause and supporting phrases.

FAQ

Will the tool modify my \cite{}, \ref{}, \label{} or equations?

No. The skill never alters citations, labels, or content inside math environments; all suggestions are emitted as commented diffs.

Can the skill run compilation and clean commands automatically?

Yes, but only when you explicitly request compilation or destructive clean options; the skill will present the exact command and ask for confirmation before execution.

How does de-AIfication work without changing facts or introducing new data?

De-AIfication rewrites visible prose for naturalness, removes empty buzzwords, and annotates claims that need evidence—without adding facts, figures, or new citations.