Transform messy documents into professionally formatted Word/PDF files with intelligent layout and auto-generated images.
When to Use
- User provides a Word document (.docx) or text file needing cleanup
- Request to "format", "beautify", "reorganize" a document
- Need to add visuals/charts to plain text content
- Converting notes/drafts into professional documents
Workflow
1. Analyze Document → 2. User Chooses Level → 3. Format → 4. Generate Images → 5. Output
Step 1: Analyze Document
Read the input file and detect:
# Content type detection keywords
CONTENT_TYPES = {
"meeting_notes": ["会议", "议程", "参会", "决议", "行动项", "meeting", "agenda"],
"report": ["报告", "总结", "分析", "数据", "report", "analysis", "summary"],
"study_notes": ["笔记", "学习", "知识点", "概念", "notes", "learning"],
"proposal": ["方案", "计划", "目标", "预算", "proposal", "plan", "budget"]
}
# Style mapping
STYLE_MAP = {
"meeting_notes": "business_formal",
"report": "business_formal",
"study_notes": "modern_minimal",
"proposal": "academic"
}
Present analysis to user:
检测结果:
- 文档类型: [类型]
- 建议风格: [风格]
- 字数: [X] 字
- 检测到数据: [是/否]
Step 2: User Chooses Processing Level
Ask user to select:
| Level | Description |
|---|---|
| Formatting only | Keep the original content, only adjust the formatting style |
| Light restructuring | Add subheadings, adjust paragraphs, add transitions |
| Deep restructuring | Can rewrite content, optimize expression, add logic |
Step 3: Apply Formatting
Style Definitions
Business Formal:
BUSINESS_STYLE = {
"title_font": "Microsoft YaHei",
"title_size": 22,
"heading1_size": 16,
"heading2_size": 14,
"body_size": 11,
"line_spacing": 1.5,
"colors": {
"primary": "#1a365d", # Deep blue
"accent": "#2b6cb0", # Medium blue
"text": "#2d3748" # Dark gray
}
}
Modern Minimal:
MODERN_STYLE = {
"title_font": "PingFang SC",
"title_size": 24,
"heading1_size": 18,
"heading2_size": 14,
"body_size": 11,
"line_spacing": 1.8,
"colors": {
"primary": "#1a202c", # Near black
"accent": "#4a5568", # Gray
"text": "#2d3748"
}
}
Academic:
ACADEMIC_STYLE = {
"title_font": "SimSun",
"title_size": 18,
"heading1_size": 15,
"heading2_size": 13,
"body_size": 12,
"line_spacing": 1.5,
"colors": {
"primary": "#1a202c",
"accent": "#4a5568",
"text": "#000000"
}
}
Document Structure
def structure_document(content, level):
"""
Structure document based on processing level
"""
if level == "format_only":
# Keep original structure, apply styles
return apply_styles(content)
elif level == "light_restructure":
# Add headings, reorder paragraphs, add transitions
sections = detect_sections(content)
sections = add_subheadings(sections)
sections = add_transitions(sections)
return apply_styles(sections)
elif level == "deep_restructure":
# Rewrite unclear parts, add logic, improve flow
sections = detect_sections(content)
sections = rewrite_unclear(sections)
sections = add_missing_logic(sections)
sections = optimize_expression(sections)
return apply_styles(sections)
Step 4: Generate Images
Image Decision Logic
def decide_images(content):
"""
Decide what images to generate based on content
"""
images = []
# Check for data → Charts
if contains_data(content):
data_sections = extract_data(content)
for section in data_sections:
chart_type = suggest_chart_type(section)
images.append({
"type": "chart",
"chart_type": chart_type,
"data": section
})
# Check for concepts → AI illustrations
concepts = extract_key_concepts(content)
if concepts:
images.append({
"type": "illustration",
"concept": concepts[0],
"style": get_document_style()
})
# Add cover image
images.append({
"type": "cover",
"title": get_document_title(content),
"style": get_document_style()
})
return images
Chart Generation
import matplotlib.pyplot as plt
import matplotlib
matplotlib.rcParams['font.sans-serif'] = ['PingFang SC', 'Microsoft YaHei']
def generate_chart(data, chart_type, style):
"""
Generate chart matching document style
"""
colors = style["colors"]
fig, ax = plt.subplots(figsize=(10, 6))
if chart_type == "bar":
ax.bar(data["labels"], data["values"], color=colors["primary"])
elif chart_type == "pie":
ax.pie(data["values"], labels=data["labels"],
colors=[colors["primary"], colors["accent"], "#e2e8f0"])
elif chart_type == "line":
ax.plot(data["x"], data["y"], color=colors["primary"], linewidth=2)
ax.set_title(data["title"], fontsize=14, color=colors["text"])
return fig
AI Illustration Generation
Use the generate-image skill:
Prompt template for business style:
"Professional business illustration of [concept],
flat design, corporate blue color scheme,
minimalist, clean background, vector style"
Prompt template for modern style:
"Modern minimalist illustration of [concept],
geometric shapes, black and white with accent color,
clean lines, abstract, professional"
Prompt template for academic style:
"Educational diagram illustrating [concept],
clean and simple, textbook style,
labeled components, neutral colors"
Image Confirmation
Present image plan to user:
配图方案:
1. 封面图: [主题描述]
2. 数据图表: [图表类型] - [数据来源段落]
3. 概念插图: [概念名称]
请选择:
A) 全部生成
B) 部分修改 (请指定)
C) 跳过配图
Step 5: Output Files
Generate Word Document
from docx import Document
from docx.shared import Pt, Inches, RGBColor
from docx.enum.text import WD_ALIGN_PARAGRAPH
def create_word_document(content, images, style):
doc = Document()
# Set page margins
for section in doc.sections:
section.top_margin = Inches(1)
section.bottom_margin = Inches(1)
section.left_margin = Inches(1.25)
section.right_margin = Inches(1.25)
# Add cover image if exists
if images.get("cover"):
doc.add_picture(images["cover"], width=Inches(6))
doc.add_page_break()
# Add content with styles
for element in content:
if element["type"] == "title":
p = doc.add_heading(element["text"], level=0)
p.alignment = WD_ALIGN_PARAGRAPH.CENTER
elif element["type"] == "heading1":
doc.add_heading(element["text"], level=1)
elif element["type"] == "heading2":
doc.add_heading(element["text"], level=2)
elif element["type"] == "paragraph":
p = doc.add_paragraph(element["text"])
p.paragraph_format.line_spacing = style["line_spacing"]
elif element["type"] == "image":
doc.add_picture(element["path"], width=Inches(5))
return doc
Convert to PDF
# Using LibreOffice (cross-platform)
libreoffice --headless --convert-to pdf --outdir [output_dir] [docx_file]
# Or using docx2pdf (requires MS Word on Windows/Mac)
# pip install docx2pdf
from docx2pdf import convert
def export_pdf(docx_path):
pdf_path = docx_path.replace('.docx', '.pdf')
convert(docx_path, pdf_path)
return pdf_path
Output Naming
def get_output_paths(input_path):
base = os.path.splitext(input_path)[0]
return {
"docx": f"{base}_formatted.docx",
"pdf": f"{base}_formatted.pdf"
}
Complete Interaction Flow
User: /word-formatter /path/to/messy_document.docx
Claude: 正在分析文档...
检测结果:
- 文档类型: 工作报告
- 建议风格: 商务正式
- 字数: 2,350 字
- 检测到数据: 是 (3处数据可视化)
请选择处理程度:
A) 仅排版 - 保持原文，调整格式
B) 轻度重组 - 添加小标题，优化结构
C) 深度重构 - 可重写内容，优化表达
User: B
Claude: 好的，将进行轻度重组。正在处理...
配图方案:
1. 封面图: 商务风格的工作报告封面
2. 柱状图: 季度销售数据对比
3. 饼图: 市场份额分布
4. 概念插图: 增长策略示意图
请选择:
A) 全部生成
B) 部分修改
C) 跳过配图
User: A
Claude: 正在生成配图...
[生成图片中...]
完成！输出文件:
- /path/to/messy_document_formatted.docx
- /path/to/messy_document_formatted.pdf
需要调整吗？
Dependencies
# Required Python packages
pip install python-docx matplotlib docx2pdf pillow
# For PDF conversion (alternative to docx2pdf)
# macOS: brew install --cask libreoffice
# Ubuntu: sudo apt install libreoffice
Error Handling
| Error | Handling method |
|---|---|
| File does not exist | Prompt the user to check the path |
| Unsupported format | Only .docx, .txt, .md are supported |
| Image generation failed | Skip this image and continue processing |
| PDF conversion failed | Output Word only, prompt to install dependencies |
Tips
- The more structured the input document, the better the formatting result
- In deep restructuring mode, Claude will try to improve the document logic
- Images will automatically match the document style's color scheme
- You can call it multiple times for iterative optimization