document-processing

文档处理

**来源：**此技能改编自 Anthropic 的文档处理技能 面向 Claude Code 和 AI Agent 的文档处理技能（pdf、docx、pptx、xlsx）。

创建、编辑和分析办公文档，包括 PDF、Word 文档、PowerPoint 演示文稿和 Excel 电子表格。

---

快速参考：选择要使用的工具

任务	文档类型	最佳工具
提取文本	PDF	pdfplumber、pdftotext
合并/拆分	PDF	pypdf、qpdf
填写表单	PDF	pdf-lib（JS）、pypdf
创建新文档	PDF	reportlab
对扫描件进行 OCR	PDF	pytesseract + pdf2image
提取文本	DOCX	pandoc、markitdown
创建新文档	DOCX	docx-js（JS）
编辑现有文档	DOCX	OOXML（解包/编辑/打包）
提取文本	PPTX	markitdown
创建新文档	PPTX	html2pptx、PptxGenJS
编辑现有文档	PPTX	OOXML（解包/编辑/打包）
数据分析	XLSX	pandas
公式/格式设置	XLSX	openpyxl

---

PDF 处理

提取文本

import pdfplumber

# Extract text with layout preservation
with pdfplumber.open("document.pdf") as pdf:
    for page in pdf.pages:
        text = page.extract_text()
        print(text)

提取表格

import pdfplumber
import pandas as pd

with pdfplumber.open("document.pdf") as pdf:
    all_tables = []
    for page in pdf.pages:
        tables = page.extract_tables()
        for table in tables:
            if table:
                df = pd.DataFrame(table[1:], columns=table[0])
                all_tables.append(df)

# Combine all tables
if all_tables:
    combined_df = pd.concat(all_tables, ignore_index=True)
    combined_df.to_excel("extracted_tables.xlsx", index=False)

合并 PDF

from pypdf import PdfWriter, PdfReader

writer = PdfWriter()
for pdf_file in ["doc1.pdf", "doc2.pdf", "doc3.pdf"]:
    reader = PdfReader(pdf_file)
    for page in reader.pages:
        writer.add_page(page)

with open("merged.pdf", "wb") as output:
    writer.write(output)

拆分 PDF

from pypdf import PdfReader, PdfWriter

reader = PdfReader("input.pdf")
for i, page in enumerate(reader.pages):
    writer = PdfWriter()
    writer.add_page(page)
    with open(f"page_{i+1}.pdf", "wb") as output:
        writer.write(output)

旋转页面

from pypdf import PdfReader, PdfWriter

reader = PdfReader("input.pdf")
writer = PdfWriter()

page = reader.pages[0]
page.rotate(90)  # Rotate 90 degrees clockwise
writer.add_page(page)

with open("rotated.pdf", "wb") as output:
    writer.write(output)

对扫描版 PDF 进行 OCR

# Requires: pip install pytesseract pdf2image
import pytesseract
from pdf2image import convert_from_path

# Convert PDF to images
images = convert_from_path('scanned.pdf')

# OCR each page
text = ""
for i, image in enumerate(images):
    text += f"Page {i+1}:\n"
    text += pytesseract.image_to_string(image)
    text += "\n\n"

print(text)

添加水印

from pypdf import PdfReader, PdfWriter

watermark = PdfReader("watermark.pdf").pages[0]
reader = PdfReader("document.pdf")
writer = PdfWriter()

for page in reader.pages:
    page.merge_page(watermark)
    writer.add_page(page)

with open("watermarked.pdf", "wb") as output:
    writer.write(output)

密码保护

from pypdf import PdfReader, PdfWriter

reader = PdfReader("input.pdf")
writer = PdfWriter()

for page in reader.pages:
    writer.add_page(page)

writer.encrypt("userpassword", "ownerpassword")

with open("encrypted.pdf", "wb") as output:
    writer.write(output)

使用 ReportLab 创建 PDF

from reportlab.lib.pagesizes import letter
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, PageBreak
from reportlab.lib.styles import getSampleStyleSheet

doc = SimpleDocTemplate("report.pdf", pagesize=letter)
styles = getSampleStyleSheet()
story = []

# Add content
title = Paragraph("Report Title", styles['Title'])
story.append(title)
story.append(Spacer(1, 12))

body = Paragraph("This is the body of the report. " * 20, styles['Normal'])
story.append(body)
story.append(PageBreak())

# Page 2
story.append(Paragraph("Page 2", styles['Heading1']))
story.append(Paragraph("Content for page 2", styles['Normal']))

doc.build(story)

命令行工具

# Extract text (poppler-utils)
pdftotext input.pdf output.txt
pdftotext -layout input.pdf output.txt  # Preserve layout

# Merge PDFs (qpdf)
qpdf --empty --pages file1.pdf file2.pdf -- merged.pdf

# Split pages
qpdf input.pdf --pages . 1-5 -- pages1-5.pdf

# Rotate pages
qpdf input.pdf output.pdf --rotate=+90:1

# Remove password
qpdf --password=mypassword --decrypt encrypted.pdf decrypted.pdf

# Extract images
pdfimages -j input.pdf output_prefix

---

Word 文档（DOCX）处理

提取文本

# Convert to markdown with pandoc
pandoc document.docx -o output.md

# With tracked changes preserved
pandoc --track-changes=all document.docx -o output.md

创建新文档（docx-js）

import { Document, Paragraph, TextRun, HeadingLevel, Packer } from 'docx';
import * as fs from 'fs';

const doc = new Document({
  sections: [{
    properties: {},
    children: [
      new Paragraph({
        text: "Document Title",
        heading: HeadingLevel.HEADING_1,
      }),
      new Paragraph({
        children: [
          new TextRun("This is a "),
          new TextRun({
            text: "bold",
            bold: true,
          }),
          new TextRun(" word in a paragraph."),
        ],
      }),
      new Paragraph({
        text: "This is another paragraph.",
      }),
    ],
  }],
});

// Export to file
const buffer = await Packer.toBuffer(doc);
fs.writeFileSync("output.docx", buffer);

创建包含表格的文档

import { Document, Paragraph, Table, TableRow, TableCell, Packer } from 'docx';

const table = new Table({
  rows: [
    new TableRow({
      children: [
        new TableCell({ children: [new Paragraph("Header 1")] }),
        new TableCell({ children: [new Paragraph("Header 2")] }),
        new TableCell({ children: [new Paragraph("Header 3")] }),
      ],
    }),
    new TableRow({
      children: [
        new TableCell({ children: [new Paragraph("Cell 1")] }),
        new TableCell({ children: [new Paragraph("Cell 2")] }),
        new TableCell({ children: [new Paragraph("Cell 3")] }),
      ],
    }),
  ],
});

const doc = new Document({
  sections: [{
    children: [
      new Paragraph({ text: "Table Example", heading: HeadingLevel.HEADING_1 }),
      table,
    ],
  }],
});

编辑现有文档（OOXML）

对于复杂编辑，请直接处理原始 OOXML：

1. 解包文档：

   python ooxml/scripts/unpack.py document.docx unpacked/
   

2. 编辑 XML 文件（主要是 word/document.xml）

3. 验证并打包：

   python ooxml/scripts/validate.py unpacked/ --original document.docx
   python ooxml/scripts/pack.py unpacked/ output.docx
   

修订工作流

使用修订功能审阅文档：

# 1. Get current state
pandoc --track-changes=all document.docx -o current.md

# 2. Unpack
python ooxml/scripts/unpack.py document.docx unpacked/

# 3. Edit using tracked change patterns
# Use <w:ins> for insertions, <w:del> for deletions

# 4. Pack final document
python ooxml/scripts/pack.py unpacked/ reviewed.docx

---

PowerPoint（PPTX）处理

提取文本

python -m markitdown presentation.pptx

创建新演示文稿（PptxGenJS）

import PptxGenJS from 'pptxgenjs';

const pptx = new PptxGenJS();

// Slide 1 - Title
const slide1 = pptx.addSlide();
slide1.addText("Presentation Title", {
  x: 1, y: 2, w: 8, h: 1.5,
  fontSize: 36,
  bold: true,
  color: "363636",
  align: "center",
});
slide1.addText("Subtitle goes here", {
  x: 1, y: 3.5, w: 8, h: 0.5,
  fontSize: 18,
  color: "666666",
  align: "center",
});

// Slide 2 - Content
const slide2 = pptx.addSlide();
slide2.addText("Key Points", {
  x: 0.5, y: 0.5, w: 9, h: 0.8,
  fontSize: 28,
  bold: true,
});
slide2.addText([
  { text: "• First important point\n", options: { bullet: true } },
  { text: "• Second important point\n", options: { bullet: true } },
  { text: "• Third important point\n", options: { bullet: true } },
], {
  x: 0.5, y: 1.5, w: 9, h: 3,
  fontSize: 18,
});

// Slide 3 - Chart
const slide3 = pptx.addSlide();
slide3.addChart(pptx.ChartType.bar, [
  { name: "Q1", labels: ["Jan", "Feb", "Mar"], values: [100, 200, 300] },
  { name: "Q2", labels: ["Apr", "May", "Jun"], values: [150, 250, 350] },
], {
  x: 1, y: 1, w: 8, h: 4,
  showLegend: true,
  legendPos: "b",
});

// Save
pptx.writeFile("output.pptx");

编辑现有演示文稿（OOXML）

# 1. Unpack
python ooxml/scripts/unpack.py presentation.pptx unpacked/

# 2. Key files:
# - ppt/slides/slide1.xml, slide2.xml, etc.
# - ppt/notesSlides/ for speaker notes
# - ppt/theme/ for styling

# 3. Validate and pack
python ooxml/scripts/validate.py unpacked/ --original presentation.pptx
python ooxml/scripts/pack.py unpacked/ output.pptx

创建缩略图网格

# Create visual overview of all slides
python scripts/thumbnail.py presentation.pptx --cols 4

将幻灯片转换为图像

# Convert to PDF first
soffice --headless --convert-to pdf presentation.pptx

# Then PDF to images
pdftoppm -jpeg -r 150 presentation.pdf slide
# Creates slide-1.jpg, slide-2.jpg, etc.

---

Excel（XLSX）处理

使用 Pandas 进行数据分析

import pandas as pd

# Read Excel
df = pd.read_excel('file.xlsx')  # Default: first sheet
all_sheets = pd.read_excel('file.xlsx', sheet_name=None)  # All sheets as dict

# Analyze
df.head()      # Preview data
df.info()      # Column info
df.describe()  # Statistics

# Filter and transform
filtered = df[df['Sales'] > 1000]
grouped = df.groupby('Category')['Revenue'].sum()

# Write Excel
df.to_excel('output.xlsx', index=False)

使用公式创建 Excel 文件（openpyxl）

from openpyxl import Workbook
from openpyxl.styles import Font, PatternFill, Alignment

wb = Workbook()
sheet = wb.active

# Add data
sheet['A1'] = 'Product'
sheet['B1'] = 'Price'
sheet['C1'] = 'Quantity'
sheet['D1'] = 'Total'

# Header formatting
for cell in ['A1', 'B1', 'C1', 'D1']:
    sheet[cell].font = Font(bold=True, color='FFFFFF')
    sheet[cell].fill = PatternFill('solid', start_color='4472C4')
    sheet[cell].alignment = Alignment(horizontal='center')

# Add data rows
data = [
    ('Widget A', 10.00, 5),
    ('Widget B', 15.00, 3),
    ('Widget C', 20.00, 8),
]

for row_idx, (product, price, qty) in enumerate(data, start=2):
    sheet[f'A{row_idx}'] = product
    sheet[f'B{row_idx}'] = price
    sheet[f'C{row_idx}'] = qty
    # FORMULA - not hardcoded value!
    sheet[f'D{row_idx}'] = f'=B{row_idx}*C{row_idx}'

# Add sum formula at bottom
last_row = len(data) + 2
sheet[f'D{last_row}'] = f'=SUM(D2:D{last_row-1})'

# Column width
sheet.column_dimensions['A'].width = 15
sheet.column_dimensions['B'].width = 10
sheet.column_dimensions['C'].width = 10
sheet.column_dimensions['D'].width = 10

wb.save('output.xlsx')

财务模型标准

颜色编码

from openpyxl.styles import Font

# Industry-standard colors
BLUE = Font(color='0000FF')   # Hardcoded inputs
BLACK = Font(color='000000')  # Formulas
GREEN = Font(color='008000')  # Links from other sheets
RED = Font(color='FF0000')    # External links

# Apply to cells
sheet['B5'].font = BLUE   # User input
sheet['B6'].font = BLACK  # Formula

数字格式

# Currency with thousands separator
sheet['B5'].number_format = '$#,##0'

# Percentage with one decimal
sheet['B6'].number_format = '0.0%'

# Zeros as dashes
sheet['B7'].number_format = '$#,##0;($#,##0);"-"'

# Multiples
sheet['B8'].number_format = '0.0x'

关键：使用公式，不要使用硬编码值

# ❌ WRONG - Hardcoding calculated values
total = df['Sales'].sum()
sheet['B10'] = total  # Hardcodes 5000

# ✅ CORRECT - Use Excel formulas
sheet['B10'] = '=SUM(B2:B9)'

# ❌ WRONG - Computing in Python
growth = (current - previous) / previous
sheet['C5'] = growth

# ✅ CORRECT - Excel formula
sheet['C5'] = '=(C4-C2)/C2'

编辑现有 Excel 文件

from openpyxl import load_workbook

# Load with formulas preserved
wb = load_workbook('existing.xlsx')
sheet = wb.active

# Modify cells
sheet['A1'] = 'New Value'
sheet.insert_rows(2)
sheet.delete_cols(3)

# Add new sheet
new_sheet = wb.create_sheet('Analysis')
new_sheet['A1'] = 'Data'

wb.save('modified.xlsx')

重新计算公式

使用公式创建或修改 Excel 文件后：

# Recalculate all formulas using LibreOffice
python recalc.py output.xlsx

---

依赖项

按需安装：

# PDF
pip install pypdf pdfplumber reportlab pytesseract pdf2image

# DOCX
npm install -g docx
pip install "markitdown[docx]"

# PPTX
npm install -g pptxgenjs
pip install "markitdown[pptx]"

# XLSX
pip install pandas openpyxl

# Command line tools
sudo apt-get install poppler-utils qpdf libreoffice pandoc

---

任务快速参考

我想要……	命令/代码
提取 PDF 文本	pdfplumber.open(f).pages[0].extract_text()
合并 PDF	pypdf.PdfWriter() + 循环
拆分 PDF	每页使用一个 PdfWriter()
对扫描版 PDF 进行 OCR	pdf2image → pytesseract
将 DOCX 转换为 MD	pandoc doc.docx -o doc.md
创建 DOCX	docx-js（JavaScript）
提取 PPTX 文本	python -m markitdown pres.pptx
创建 PPTX	PptxGenJS（JavaScript）
分析 Excel	pandas.read_excel()
使用公式的 Excel 文件	openpyxl

---

致谢与署名

此技能基于 Anthropic 的出色成果。

原始仓库：https://github.com/anthropics/skills/tree/main/skills/document-processing

特别感谢 Anthropic 慷慨地贡献开源成果，助力打造了这一技能集。 由 webconsulting.at 改编，供此技能集使用