Workwork - Academic Review Writing & Formatting Assistant
Purpose
This skill provides comprehensive tools for academic review writing and formatting, specifically optimized for Chinese academic standards. It automates reference checking, format validation, document generation, and quality control processes to help researchers produce compliant, professional academic papers efficiently.
When to Use This Skill
Use this skill when:
- - Writing or formatting academic literature reviews
- Checking reference list formats and citations
- Validating document structure according to Chinese academic standards
- Generating Word documents with proper formatting
- Filtering references by journal quality (e.g., removing non-core journals)
- Fixing reference numbering and citation consistency
- Proofreading for typos and grammar errors
Core Capabilities
This skill provides ten functional modules accessible through scripts in the scripts/ directory:
1. Literature Integrity Checker ⭐ NEW (CRITICAL)
- - Script: INLINECODE1
- Purpose: Comprehensive validation of reference integrity and citation consistency
- What it checks:
- Reference completeness (authors, title, journal, year, pages)
- Citation consistency (all citations have corresponding references)
- Reference numbering continuity
- Duplicate references
- Citation format compliance
- Invalid/orphan citations
-
Consecutive duplicate citations 🆕 NEW: Detects when the same reference is cited in 3+ consecutive sentences
- - Critical Feature: Returns non-zero exit code on critical errors, preventing submission with errors
- Auto-Open Feature 🆕 NEW: Automatically opens the final Word document (
your_review_上标版.docx) after successful check (no critical issues) - Usage:
# Standard check with auto-open enabled (default)
python scripts/literature_integrity_checker.py your_review.md
# With deep verification
python scripts/literature_integrity_checker.py your_review.md --deep-check
# Disable auto-open
python scripts/literature_integrity_checker.py your_review.md --no-auto-open
- - Output: INLINECODE3
- Note: When check passes (no critical issues), the script automatically searches for and opens
your_review_上标版.docx or INLINECODE5
2. Unified Checker (RECOMMENDED - Most Efficient)
- - Script: INLINECODE6
- Purpose: Run all checks at once in a single operation
- What it does:
- Reference format validation
- Citation accuracy checking
- Typo and grammar detection
- Document format compliance checking
python scripts/unified_checker.py your_review.md
3. Reference Formatter
- - Script: INLINECODE8
- Purpose: Validate and format reference citations and bibliography
- What it checks:
- Citation marker positioning (before punctuation)
- Author name formatting (adding "等" when exceeding limit)
- Reference numbering continuity
- Duplicate reference numbers
- Figure/table citation markers
- - Configuration: Edit
templates/ref_format_default.yml for custom rules - Usage:
CODEBLOCK2
4. Typo & Grammar Checker
- - Script: INLINECODE10
- Purpose: Detect common typos, grammar errors, punctuation issues
- What it checks:
- Common Chinese typos
- Mismatched quotes/brackets
- Punctuation usage
- Number formatting
CODEBLOCK3
5. Reference Accuracy Checker
- - Script: INLINECODE11
- Purpose: Verify citation validity and reference integrity
- What it checks:
- Invalid citations (referencing non-existent references)
- Orphaned references (not cited in text)
- Reference numbering continuity
- Citation numbering in text
- Citation marker positioning
CODEBLOCK4
6. Document Format Checker
- - Script: INLINECODE12
- Purpose: Validate document structure and formatting
- What it checks:
- Main title presence and formatting
- Abstract and keywords sections
- Chapter structure and numbering
- References section
- Paragraph formatting
- Heading hierarchy
- List formatting
CODEBLOCK5
7. Reference Filter
- - Script: INLINECODE13
- Purpose: Filter references based on journal quality lists
- What it does: Identifies and removes references from non-core journals
- Setup required: Edit the script's journal lists before running
- Usage:
CODEBLOCK6
8. Reference Number Fixer
- - Script: INLINECODE14
- Purpose: Renumber references and update all citations in text
- What it does:
- Extracts all references
- Renumbers them sequentially
- Updates all citation markers in text
- Removes invalid citations
- - Warning: Directly modifies the source file—backup first
- Usage:
CODEBLOCK7
9. Word Document Generator
- - Script:
create_word_doc_v3.js (Node.js) - Purpose: Generate Word documents with Chinese academic formatting
- What it produces:
- Font: SimSun (宋体) for Chinese, Times New Roman for English
- Font size: 小四 for body, 三号 for headings
- Line spacing: 1.5倍
- Margins: 2.54cm
- - Requirements: Node.js 14+,
docx npm package - Usage:
CODEBLOCK8
10. Word Document with Superscript Citations
- - Script:
create_word_with_superscript.js (Node.js) - Purpose: Generate Word documents with superscript reference citations (右上标格式)
- Important: Citation format is
[n] (with brackets) in superscript - e.g., [1], INLINECODE20 - Features:
-
Smart Font Handling: Chinese text uses SimSun (宋体), English/numbers/punctuation use Times New Roman
-
Superscript Citations: Body text citations appear as superscript:
[¹],
[²⁻³]
-
Italic Scientific Names: Automatically detects and italicizes species Latin names (e.g.,
Rhinopithecus bieti)
-
Font Size: Body text and references use 小四 (12pt), headings use appropriate sizes
-
Auto-Open: Automatically opens the generated document after creation
- Body text citations appear as superscript:
[¹],
[²⁻³]
- Reference list entries remain normal format (not superscript)
- All other formatting same as standard generator
- - Requirements: Node.js 14+,
docx npm package - Usage:
node scripts/create_word_with_superscript.js your_review.md
- - Output:
your_review_上标版.docx (auto-opens after generation)
11. Citation Pattern Analyzer ⭐ NEW
- - Script: INLINECODE27
- Purpose: Detect "whole-paragraph citation split into per-sentence annotations" pattern
- What it checks:
- Same reference cited in 2+ consecutive sentences within a paragraph
- "Serious" flag for 3+ consecutive sentences
- "Warning" flag for 2 consecutive sentences
- Distinguishes valid total-then-detail structures from unnecessary repetition
- - Academic rule: When an entire paragraph derives from one source, cite only at the first or last sentence, not every sentence
- Usage:
python scripts/analyze_citation_pattern.py your_review.md
For most use cases, follow this 9-step workflow:
Step 1: Write Review
Create your academic review in Markdown format: INLINECODE29
Step 2: Run Literature Integrity Check ⭐ CRITICAL
这是最重要的检查步骤! 运行文献完整性检查,确保所有引用和文献都正确无误:
python scripts/literature_integrity_checker.py your_review.md
This takes ~2 minutes and generates INLINECODE30
⚠️ 重要: 此检查会发现以下严重错误:
- - 无效引用(引用编号超出参考文献范围)
- 孤立引用(正文中有引用但参考文献列表中不存在)
- 参考文献编号不连续
- 重复的参考文献
- 文献信息不完整(缺少作者、标题、期刊、年份等)
如果发现严重问题,脚本会返回错误码,阻止继续操作!
Step 3: Run Unified Check
Execute unified checker to run all validations at once:
python scripts/unified_checker.py your_review.md
This takes ~5 minutes and generates INLINECODE31
Step 4: Review Report
Read the unified report:
cat your_review_unified_report.md
The report contains:
- - Problem statistics (total/critical/warnings/info)
- Detailed issues by category
- Fix recommendations
Step 5: Fix Issues
Edit
your_review.md based on report findings:
- - Fix critical issues (🔴 must fix) first
- Then warnings (🟡 should fix)
- Finally info items (🔵 optional)
Step 6: Filter References (Optional)
If you need to remove non-core journals:
- 1. Edit
scripts/filter_references.py to set journal lists - Run: INLINECODE34
Step 7: Fix Reference Numbers
If you removed references, renumber them:
python scripts/extract_and_fix_references.py
Warning: This modifies the source file directly—backup first!
Step 8: Final Verification
Quick verification:
CODEBLOCK15
再次运行文献完整性检查(推荐):
CODEBLOCK16
✨ 新功能: 如果检查通过(无严重问题),脚本会自动打开 your_review_上标版.docx 文档!
Step 9: Generate Word Document
Create the final formatted Word document:
Standard format (inline citations):
CODEBLOCK17
Superscript format (右上标引用) - RECOMMENDED for Chinese academic papers:
node scripts/create_word_with_superscript.js your_review.md
⚠️
Important:
- - Superscript citations include brackets:
[1], [2-3] appear as superscript - Automatically handles mixed Chinese/English fonts (宋体 for Chinese, Times New Roman for English/numbers)
- Auto-italicizes scientific names (e.g., Rhinopithecus bieti)
- Document auto-opens after generation
Output: your_review_上标版.docx
🚀 快速工作流(推荐)
对于新的文献综述,使用以下快速工作流:
CODEBLOCK19
Alternative Workflows
Complete Workflow (Step-by-Step)
Use when you need detailed information about each check type. Run each checker individually:
- 1. INLINECODE39
- INLINECODE40
- INLINECODE41
- INLINECODE42
View each report separately for detailed analysis.
Quick Fix Workflow
Use when you've only made minor edits after previous checks. Run only the relevant checker:
- - Fixed citations? → INLINECODE43
- Fixed typos? → INLINECODE44
- Fixed formatting? → INLINECODE45
- Fixed reference format? → INLINECODE46
Important Notes
Avoid Redundant Checks
- - ❌ Wrong: Run unified check → fix → run unified check again → fix again...
- ✅ Right: Run unified check → fix → targeted recheck → generate Word
The unified checker is designed to be run once per major editing session. After fixing issues, only run the specific checkers relevant to what you changed.
File Backup
- - Scripts that modify source files:
extract_and_fix_references.py, INLINECODE48 - Always backup before running:
CODEBLOCK20
Configuration Files
- - Main config: INLINECODE49
- Edit this file to customize:
- Citation style (superscript vs inline)
- Max authors before adding "等"
- Journal filtering rules
- Quality check options
Dependencies
- - Python 3.7+
- Node.js 14+
- Python packages:
pyyaml, python-docx (optional) - Node.js package: INLINECODE52
Example Use Cases
Case 1: First Draft - Full Check
CODEBLOCK21
Case 2: Reference Numbering Fix
CODEBLOCK22
Case 3: Minor Edits - Quick Fix
CODEBLOCK23
Troubleshooting
Python encoding issues on Windows
If you encounter GBK vs UTF-8 encoding errors:
- - Ensure scripts are saved with UTF-8 encoding
- Use
encoding='utf-8' in file operations
Node.js docx package not found
CODEBLOCK24
Missing YAML configuration
Copy
templates/ref_format_default.yml to your working directory and edit it.
Additional Resources
See references/ directory for:
- - Detailed workflow guides
- Configuration examples
- Troubleshooting tips
- Feature checklists
See scripts directory for:
- - All executable checkers and formatters
- Helper utilities for document analysis
- Validation and repair tools
Version History
- - v3.3.0 (2026-03-20): 功能更新 - 新增引用模式分析器 + 深度引用审查
-
新增:
analyze_citation_pattern.py - 专项检测整段引用被拆分逐句标注的问题
- 检测连续句子中同一引用重复出现的模式
- 3句及以上标记为严重,2句标记为警告
- 生成详细报告含原文句子和修复建议
-
修复(典型案例): 学术文献综述中发现的严重引用错误类型:
- 引用编号指向错误文献(如用贻贝实验引用编号标注灵长类研究)→ 核查并更正
- 超出参考文献范围的引用编号 → 全部清理或替换为有效编号
- 引用文献内容与正文上下文不符 → 核对并修正
- 整段内容中间重复标注 → 合并为段首+段末引用
-
最佳实践总结: 引用错配比重复引用更危险,需先核验参考文献实际内容
- - v3.2.0 (2026-03-19): 功能更新 - 新增自动打开最终文档功能
-
新增: 文献完整性检查器支持检查通过后自动打开 Word 文档
- 自动查找
_上标版.docx 或
.docx 文件
- 仅在无严重问题时自动打开
- 跨平台支持(Windows/macOS/Linux)
- 可通过
--no-auto-open 参数禁用
-
更新: SKILL.md 文档,添加快速工作流说明
-
改进: 优化用户体验,简化工作流程
- - v3.1.0 (2026-03-19): 功能更新 - 新增连续相同引用检查
-
新增: 连续相同引用检测功能
- 自动检测同一段落中连续三句话或更多句子引用同一篇文献的情况
- 遵循学术规范:同一段落中连续引用同一篇文献时,建议只在最后一句话保留引用
- 作为警告级别提示,不阻止文档生成
-
改进: 优化检测逻辑,只标记相邻句子中的重复引用,避免误报段落首尾的正常引用模式
-
更新: 更新文献完整性检查器功能说明和版本日志
- - v3.0.0 (2026-03-19): 重大更新 - 新增文献完整性检查系统
-
新增:
literature_integrity_checker.py - 全面的文献完整性验证
- 检查文献信息完整性(作者、标题、期刊、年份、页码)
- 验证引用与文献列表的一致性
- 检测无效引用和孤立引用
- 检查参考文献编号连续性
- 识别重复文献
-
关键特性: 发现严重错误时返回非零退出码,阻止错误提交
-
更新: 推荐工作流程,将文献完整性检查作为关键步骤
-
改进: 更严格的错误检测机制,杜绝文献引用错误
- - v2.3.0 (2026-03-19): Enhanced superscript Word generator with:
- Smart mixed font handling (Chinese SimSun + English Times New Roman)
- Auto-italicization of scientific names (Latin species names)
- Auto-open document after generation
- Body text and references use 小四 (12pt) font size
- - v2.2.0 (2026-03-19): Added superscript citation Word generator (
create_word_with_superscript.js). Important: Superscript citations retain brackets [n] format. - v2.1.0 (2026-03-19): Added unified checker, optimized workflows, eliminated redundant checks
- v2.0.0 (2026-03-19): Added typo checking, reference accuracy, document format validation
- v1.0.0 (2026-03-19): Initial version with basic reference formatting and Word generation
Workwork - 学术综述写作与格式排版助手
目的
本技能提供全面的学术综述写作与格式排版工具,专门针对中文学术标准进行了优化。它能够自动完成参考文献检查、格式验证、文档生成和质量控制流程,帮助研究人员高效地撰写符合规范的、专业的学术论文。
何时使用本技能
在以下情况下使用本技能:
- - 撰写或格式化学术文献综述
- 检查参考文献列表格式和引文
- 根据中文学术标准验证文档结构
- 生成具有正确格式的Word文档
- 根据期刊质量筛选参考文献(例如,删除非核心期刊)
- 修正参考文献编号和引文一致性
- 校对拼写和语法错误
核心功能
本技能提供十个功能模块,可通过scripts/目录中的脚本访问:
1. 文献完整性检查器 ⭐ 新增(关键)
- - 脚本: literatureintegritychecker.py
- 目的: 全面验证参考文献完整性和引文一致性
- 检查内容:
- 参考文献完整性(作者、标题、期刊、年份、页码)
- 引文一致性(所有引文都有对应的参考文献)
- 参考文献编号连续性
- 重复参考文献
- 引文格式合规性
- 无效/孤立引文
-
连续重复引文 🆕 新增:检测同一参考文献在连续3句或更多句子中被引用的情况
- - 关键特性: 发现严重错误时返回非零退出码,阻止提交错误文档
- 自动打开功能 🆕 新增:检查通过(无严重问题)后,自动打开最终的Word文档(yourreview上标版.docx)
- 使用方法:
bash
# 标准检查,自动打开功能已启用(默认)
python scripts/literature
integritychecker.py your_review.md
# 深度验证
python scripts/literatureintegritychecker.py your_review.md --deep-check
# 禁用自动打开
python scripts/literatureintegritychecker.py your_review.md --no-auto-open
- - 输出: yourreviewliteratureintegrityreport.md
- 注意: 当检查通过(无严重问题)时,脚本会自动搜索并打开yourreview上标版.docx或your_review.docx
2. 统一检查器(推荐 - 最高效)
- - 脚本: unified_checker.py
- 目的: 一次性运行所有检查
- 功能:
- 参考文献格式验证
- 引文准确性检查
- 拼写和语法检测
- 文档格式合规性检查
bash
python scripts/unified
checker.py yourreview.md
- - 输出: yourreviewunified_report.md
3. 参考文献格式化器
- - 脚本: reference_formatter.py
- 目的: 验证并格式化参考文献引文和书目
- 检查内容:
- 引文标记位置(标点符号之前)
- 作者姓名格式(超出限制时添加等)
- 参考文献编号连续性
- 重复的参考文献编号
- 图表引文标记
- - 配置: 编辑templates/refformatdefault.yml以自定义规则
- 使用方法:
bash
python scripts/reference
formatter.py yourreview.md
python scripts/reference
formatter.py yourreview.md custom_config.yml
4. 拼写与语法检查器
- - 脚本: typogrammarchecker.py
- 目的: 检测常见拼写错误、语法错误、标点问题
- 检查内容:
- 常见中文拼写错误
- 不匹配的引号/括号
- 标点符号使用
- 数字格式
bash
python scripts/typo
grammarchecker.py your_review.md
5. 参考文献准确性检查器
- - 脚本: referenceaccuracychecker.py
- 目的: 验证引文有效性和参考文献完整性
- 检查内容:
- 无效引文(引用不存在的参考文献)
- 孤立参考文献(正文中未引用)
- 参考文献编号连续性
- 正文中的引文编号
- 引文标记位置
bash
python scripts/reference
accuracychecker.py your_review.md
6. 文档格式检查器
- - 脚本: documentformatchecker.py
- 目的: 验证文档结构和格式
- 检查内容:
- 主标题存在性和格式
- 摘要和关键词部分
- 章节结构和编号
- 参考文献部分
- 段落格式
- 标题层级
- 列表格式
bash
python scripts/document
formatchecker.py your_review.md
7. 参考文献过滤器
- - 脚本: filter_references.py
- 目的: 根据期刊质量列表筛选参考文献
- 功能: 识别并删除非核心期刊的参考文献
- 需要设置: 运行前编辑脚本中的期刊列表
- 使用方法:
bash
python scripts/filter_references.py
8. 参考文献编号修复器
- - 脚本: extractandfix_references.py
- 目的: 重新编号参考文献并更新正文中的所有引文
- 功能:
- 提取所有参考文献
- 按顺序重新编号
- 更新正文中的所有引文标记
- 删除无效引文
bash
python scripts/extract
andfix_references.py
9. Word文档生成器
- - 脚本: createworddoc_v3.js (Node.js)
- 目的: 生成符合中文学术格式的Word文档
- 生成内容:
- 字体:中文使用宋体,英文使用Times New Roman
- 字号:正文小四,标题三号
- 行距:1.5倍
- 页边距:2.54cm
- - 要求: Node.js 14+, docx npm包
- 使用方法:
bash
node scripts/create
worddoc_v3.js
10. 带右上标引文的Word文档
- - 脚本: createwordwith_superscript.js (Node.js)
- 目的: 生成带有右上标参考文献引文的Word文档(右上标格式)
- 重要: 引文格式为[n](带括号)作为上标——例如,[1]、[2-3]
- 特性:
-
智能字体处理: 中文文本使用宋体,英文/数字/标点使用Times New Roman
-
上标引文: 正文引文显示为上标:[¹]、[²⁻³]
-
斜体学名: 自动检测并将物种拉丁名设为斜体(例如,
Rhinopithecus bieti)
-
字号: 正文和参考文献使用小四(12pt),标题使用适当字号
-
自动打开: 生成后自动打开文档
- 正文引文显示为上标:[¹]、[²⁻³]
- 参考文献列表条目保持正常格式(非上标)
- 所有其他格式与标准生成器相同
- - 要求: Node.js 14+, docx npm包
- 使用方法:
bash
node scripts/create
wordwith
superscript.js yourreview.md
- - 输出: yourreview上标版.docx(生成后自动打开)
11. 引文模式分析器 ⭐ 新增
- - 脚本: analyzecitationpattern.py
- 目的: 检测整段引用被拆分为逐句标注的模式
- 检查内容:
- 同一参考文献在段落中连续2句或更多句中被引用
- 连续3句及以上标记为严重
- 连续2句标记为警告
- 区分有效的总分结构与不必要的重复
- - 学术规则: 当整个段落来源于同一来源时,仅在首句或末句引用,而非每句都引用
- 使用方法:
bash
python scripts/analyze
citationpattern.py your_review.md
- - 输出: yourreviewcitationpatternreport.md
对于大多数使用场景,请遵循以下9步工作流程:
第1步:撰写综述
以Markdown格式创建您的学术综述:your_review.md
第2步:运行