项目	内容
来源	skills/pdf/SKILL.md
分类	文档 & 办公文件
安装	克隆 Skill 分支后，将 `skills/pdf/` 目录复制到 `~/.claude/skills/`（Claude Code）、`~/.trae/skills/`（Trae IDE）或 `~/.qclaw/skills/`（QClaw/OpenClaw），详见 Skill 安装方式

PDF Processing Guide

Overview

This guide covers essential PDF processing operations using Python libraries and command-line tools. For advanced features, JavaScript libraries, and detailed examples, see REFERENCE.md. If you need to fill out a PDF form, read FORMS.md and follow its instructions.

Quick Start

from pypdf import PdfReader, PdfWriter

# Read a PDF
reader = PdfReader("document.pdf")
print(f"Pages: {len(reader.pages)}")

# Extract text
text = ""
for page in reader.pages:
    text += page.extract_text()

Python Libraries

pypdf - Basic Operations

Merge PDFs

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)

Split PDF

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)

Extract Metadata

reader = PdfReader("document.pdf")
meta = reader.metadata
print(f"Title: {meta.title}")
print(f"Author: {meta.author}")
print(f"Subject: {meta.subject}")
print(f"Creator: {meta.creator}")

Rotate Pages

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)

pdfplumber - Text and Table Extraction

Extract Text with Layout

import pdfplumber

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

Extract Tables

with pdfplumber.open("document.pdf") as pdf:
    for i, page in enumerate(pdf.pages):
        tables = page.extract_tables()
        for j, table in enumerate(tables):
            print(f"Table {j+1} on page {i+1}:")
            for row in table:
                print(row)

Advanced Table Extraction

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:  # Check if table is not empty
                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)

reportlab - Create PDFs

🚨 MANDATORY RULE — Read before writing any reportlab code Every reportlab script MUST import and call setup_chinese_pdf() first — even for English-only content. Step 0 — import the helper (this skill ships it as scripts/setup_chinese_pdf.py):
import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "scripts"))
from setup_chinese_pdf import setup_chinese_pdf
If your script is NOT in the skill root directory, use the absolute path to scripts/ instead. Then follow these rules:

Call cn_font, styles = setup_chinese_pdf() as the very first reportlab operation.

Use the returned styles for all Paragraph(...) calls — do NOT call getSampleStyleSheet() separately.

For any custom ParagraphStyle, use parent=styles['X'] so it inherits the CJK font.

For Canvas API, call c.setFont(cn_font, size) before every c.drawString() / c.drawCentredString().

NEVER use Helvetica, Times-Roman, or Courier for any text that may contain Chinese.

Skipping any of these steps causes every Chinese character to render as a solid black box. There is NO error — the script exits 0 and the PDF looks silently broken. All examples below already follow this pattern. You can use any of them as a starting template.

Basic PDF Creation

import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "scripts"))
from setup_chinese_pdf import setup_chinese_pdf

from reportlab.lib.pagesizes import letter
from reportlab.pdfgen import canvas

# MUST be the very first reportlab operation
cn_font, styles = setup_chinese_pdf()

c = canvas.Canvas("hello.pdf", pagesize=letter)
width, height = letter

# Set font before drawing text (required for both English and Chinese)
c.setFont(cn_font, 12)
c.drawString(100, height - 100, "Hello World!")
c.drawString(100, height - 120, "This is a PDF created with reportlab")

# Add a line
c.line(100, height - 140, 400, height - 140)

# Save
c.save()

Create PDF with Multiple Pages

import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "scripts"))
from setup_chinese_pdf import setup_chinese_pdf

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

# MUST be the very first reportlab operation
cn_font, styles = setup_chinese_pdf()

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

# Add content — styles from setup_chinese_pdf() already have CJK font set
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']))

# Build PDF
doc.build(story)

Chinese / CJK PDF Content

⚠️ IMPORTANT — NEVER fall back to English: If the user asked for Chinese content, the PDF must contain Chinese. Do not replace Chinese content with English and claim it is a “limitation” — that is incorrect behaviour. Follow the instructions below to produce correct Chinese output on every platform.

ReportLab’s built-in fonts (Helvetica, Times-Roman, Courier) are Latin-only — they contain no CJK glyphs. Passing Chinese text to these fonts causes characters to silently render as blank spaces or boxes. The script exits with code 0, so there is no error to catch; the PDF simply has no Chinese text. Why intermittent garbling happens — three traps in reportlab:

ParagraphStyle copies parent attributes at construction time. ParagraphStyle('X', parent=styles['Title']) copies fontName='Helvetica-Bold' into the new object’s __dict__ immediately. Patching the parent afterwards has no effect on already-created children. Every style that may contain Chinese MUST receive fontName=cn_font explicitly, or inherit from an already-patched parent.
getSampleStyleSheet() returns a new instance every call. Patching one instance does not affect the next call. So styles['Normal'].fontName = cn_font only works for that one styles object.
TableStyle FONTNAME is silently ignored for Paragraph cells. When a table cell contains a Paragraph object, the Paragraph’s own fontName wins — the TableStyle('FONTNAME', ...) setting has zero effect on it.

Fix: use setup_chinese_pdf() below. It returns (cn_font, styles) where styles is a pre-patched stylesheet. Use it for ALL text — then Chinese never breaks.

The complete solution — `setup_chinese_pdf()`

Preferred: import from the pre-built module scripts/setup_chinese_pdf.py (see MANDATORY RULE above).
Fallback: if importing is not possible (e.g. Windows base64 workflow), copy the function below verbatim into your script.
Source code: scripts/setup_chinese_pdf.py

import os
import platform
from reportlab.pdfbase import pdfmetrics
from reportlab.pdfbase.ttfonts import TTFont
from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle


def setup_chinese_pdf():
    """
    Register a system CJK font and return (cn_font, styles).

    cn_font: the registered font name — pass to canvas.setFont() and any
             ParagraphStyle you create manually via fontName=cn_font.

    styles:  a StyleSheet1 where ALL default ParagraphStyles already use
             the CJK font. Use styles['Title'], styles['Normal'], etc.
             directly — Chinese will render correctly without extra args.

    Also works for child styles: ParagraphStyle('X', parent=styles['Title'])
    will inherit the CJK fontName because the parent was patched BEFORE the
    child was constructed (ParagraphStyle copies parent attrs at init time).
    """
    system = platform.system()

    if system == 'Darwin':  # macOS
        # PingFang.ttc does NOT exist on disk — it is a virtual system font.
        # STHeiti / Songti are real files present on all macOS versions.
        candidates = [
            ('/System/Library/Fonts/STHeiti Light.ttc',       'STHeiti',       0),
            ('/System/Library/Fonts/STHeiti Medium.ttc',      'STHeitiMedium', 0),
            ('/System/Library/Fonts/Supplemental/Songti.ttc', 'Songti',        0),
            ('/Library/Fonts/Arial Unicode MS.ttf',           'ArialUnicode',  0),
        ]
    elif system == 'Windows':
        candidates = []
        dirs = []
        windir = os.environ.get('WINDIR', 'C:\\Windows')
        dirs.append(os.path.join(windir, 'Fonts'))
        local = os.environ.get('LOCALAPPDATA', '')
        if local:
            dirs.append(os.path.join(local, 'Microsoft', 'Windows', 'Fonts'))
        for d in dirs:
            for fname, name, idx in [
                ('msyh.ttc',   'MicrosoftYaHei', 0),   # 微软雅黑 — Win10/11
                ('msyhbd.ttc', 'MicrosoftYaHeiBold', 0),
                ('simhei.ttf', 'SimHei',          0),   # 黑体
                ('simsun.ttc', 'SimSun',          0),   # 宋体
                ('mingliu.ttc','MingLiU',         0),   # 細明體
            ]:
                candidates.append((os.path.join(d, fname), name, idx))
    else:  # Linux
        candidates = [
            ('/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc', 'NotoSansCJK', 0),
            ('/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc', 'NotoSansCJK', 0),
            ('/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc',           'WQYZenHei',   0),
            ('/usr/share/fonts/truetype/droid/DroidSansFallbackFull.ttf','DroidSans',  0),
        ]

    cn_font = None
    for font_path, font_name, idx in candidates:
        if os.path.exists(font_path):
            try:
                pdfmetrics.registerFont(TTFont(font_name, font_path, subfontIndex=idx))
                cn_font = font_name
                break
            except Exception:
                continue
    if cn_font is None:
        raise RuntimeError(
            f"No CJK font found on {system}.\n"
            "  macOS:   check /System/Library/Fonts/\n"
            "  Windows: ensure msyh.ttc or simhei.ttf in %WINDIR%\\Fonts\n"
            "  Linux:   sudo apt install fonts-noto-cjk"
        )

    # Build a stylesheet with ALL styles pre-patched to use the CJK font.
    # This eliminates garbling even if fontName= is accidentally omitted.
    styles = getSampleStyleSheet()
    for style in styles.byName.values():
        if isinstance(style, ParagraphStyle):
            style.fontName = cn_font

    return cn_font, styles

Using `setup_chinese_pdf()` — Platypus (Paragraph / Table)

import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "scripts"))
from setup_chinese_pdf import setup_chinese_pdf

from reportlab.lib.pagesizes import A4
from reportlab.lib.styles import ParagraphStyle
from reportlab.lib.enums import TA_CENTER
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Table, TableStyle
from reportlab.lib import colors

# ── 1. MUST be the first reportlab call ──
cn_font, styles = setup_chinese_pdf()

# styles['Title'], styles['Normal'], styles['Heading1'] etc. already use CJK font.
# Use them directly — no need to pass fontName.

# If you create custom styles, use parent=styles['X'] so they inherit the CJK font:
title_style = ParagraphStyle('CnTitle', parent=styles['Title'], fontSize=20, alignment=TA_CENTER)
body_style  = ParagraphStyle('CnBody',  parent=styles['Normal'], fontSize=11, leading=18)
# title_style.fontName is already the CJK font — inherited from patched parent.

doc = SimpleDocTemplate("report_cn.pdf", pagesize=A4)

# ── 2. Tables: wrap EVERY cell in Paragraph(text, styles['X']) ──
# TableStyle FONTNAME is silently ignored for Paragraph cells.
# ALWAYS use Paragraph(..., styles['Normal']) for any cell that contains Chinese.
table_data = [
    [Paragraph('指标',    styles['Normal']), Paragraph('2024年',   styles['Normal'])],
    [Paragraph('市场规模', styles['Normal']), Paragraph('32.1亿元', styles['Normal'])],
]
t = Table(table_data, colWidths=[200, 150])
t.setStyle(TableStyle([
    ('BACKGROUND', (0, 0), (-1, 0), colors.HexColor('#2E4057')),
    ('TEXTCOLOR',  (0, 0), (-1, 0), colors.white),
    ('GRID',       (0, 0), (-1,-1), 0.5, colors.grey),
    ('FONTNAME',   (0, 0), (-1,-1), cn_font),  # only affects plain-string cells
    ('FONTSIZE',   (0, 0), (-1,-1), 11),
]))

story = [
    Paragraph("国内市场分析报告", title_style),
    Spacer(1, 12),
    Paragraph("正文内容，中文可以正常显示。English and 中文 mixed.", body_style),
    Spacer(1, 20),
    t,
]
doc.build(story)

Using `setup_chinese_pdf()` — Canvas API

Canvas drawString / drawCentredString requires explicit c.setFont(cn_font, size) before every call that contains Chinese. The font is NOT inherited between draw calls.

import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "scripts"))
from setup_chinese_pdf import setup_chinese_pdf

from reportlab.pdfgen import canvas

cn_font, styles = setup_chinese_pdf()

c = canvas.Canvas("cn_canvas.pdf")

c.setFont(cn_font, 16)
c.drawString(100, 750, "你好，世界！Hello World！")

c.setFont(cn_font, 12)   # must set again when size changes
c.drawString(100, 720, "中文内容必须在 setFont 之后绘制")

c.save()

Windows: write Python scripts without any Chinese in the shell command

On Windows, both PowerShell and CMD interpret the command line before passing it to Python. Their default encoding (GBK / CP936) corrupts Chinese string literals in the shell command before Python ever sees them — even inside python -c "...". Do NOT do this (Chinese in the shell string — always breaks on GBK consoles):

# ❌ WRONG — PowerShell corrupts Chinese before python receives it
python -c "open('x.py','w',encoding='utf-8').write('print(\"中文\")')"

Correct approach: encode Chinese content as base64, decode inside Python The shell command contains only ASCII. Python receives the base64 string intact regardless of the console encoding, then decodes it back to UTF-8 internally.

# Step 1: on any machine that has Python, generate the base64 payload once:
#   python -c "import base64; print(base64.b64encode(open('script.py','rb').read()).decode())"
# Then paste the result as PAYLOAD below.

# Step 2: on Windows, run the following (all ASCII — no Chinese in the shell):
python -c "import base64,os; open('gen_report.py','wb').write(base64.b64decode('<PAYLOAD>'))"
python gen_report.py

Full workflow example (model should follow this pattern every time):

# 1. Build the script content and base64-encode it in one python call.
#    The python -c string here is pure ASCII — base64 payload carries the Chinese.
python -c "
import base64, textwrap

script = textwrap.dedent('''
    # -*- coding: utf-8 -*-
    import os, platform
    from reportlab.pdfbase import pdfmetrics
    from reportlab.pdfbase.ttfonts import TTFont
    from reportlab.lib.pagesizes import A4
    from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
    from reportlab.lib.enums import TA_CENTER
    from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer

    def setup_chinese_pdf():
        system = platform.system()
        if system == \"Windows\":
            candidates = []
            dirs = []
            windir = os.environ.get(\"WINDIR\", \"C:\\\\Windows\")
            dirs.append(os.path.join(windir, \"Fonts\"))
            local = os.environ.get(\"LOCALAPPDATA\", \"\")
            if local:
                dirs.append(os.path.join(local, \"Microsoft\", \"Windows\", \"Fonts\"))
            for d in dirs:
                for fname, name, idx in [
                    (\"msyh.ttc\",\"MicrosoftYaHei\",0),
                    (\"simhei.ttf\",\"SimHei\",0),
                    (\"simsun.ttc\",\"SimSun\",0),
                ]:
                    candidates.append((os.path.join(d, fname), name, idx))
        elif system == \"Darwin\":
            candidates = [
                (\"/System/Library/Fonts/STHeiti Light.ttc\", \"STHeiti\", 0),
                (\"/System/Library/Fonts/STHeiti Medium.ttc\", \"STHeitiMedium\", 0),
            ]
        else:
            candidates = [
                (\"/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc\", \"NotoSansCJK\", 0),
                (\"/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc\", \"WQYZenHei\", 0),
            ]
        cn_font = None
        for font_path, font_name, idx in candidates:
            if os.path.exists(font_path):
                try:
                    pdfmetrics.registerFont(TTFont(font_name, font_path, subfontIndex=idx))
                    cn_font = font_name
                    break
                except Exception:
                    continue
        if cn_font is None:
            raise RuntimeError(\"No CJK font found\")
        styles = getSampleStyleSheet()
        for style in styles.byName.values():
            if isinstance(style, ParagraphStyle):
                style.fontName = cn_font
        return cn_font, styles

    cn_font, styles = setup_chinese_pdf()
    title_style = ParagraphStyle(\"T\", parent=styles[\"Title\"], fontSize=20, alignment=TA_CENTER)
    body_style  = ParagraphStyle(\"B\", parent=styles[\"Normal\"], fontSize=11, leading=18)
    out = os.path.join(os.path.expanduser(\"~\"), \"Desktop\", \"report_cn.pdf\")
    doc = SimpleDocTemplate(out)
    doc.build([
        Paragraph(\"\u56fd\u5185\u5e02\u573a\u5206\u6790\u62a5\u544a\", title_style),
        Spacer(1, 12),
        Paragraph(\"\u8fd9\u662f\u6b63\u6587\u5185\u5bb9\uff0c\u4e2d\u6587\u53ef\u4ee5\u6b63\u5e38\u663e\u793a\u3002\", body_style),
    ])
    print(\"PDF saved:\", out)
''').encode('utf-8')

payload = base64.b64encode(script).decode()
# Write a pure-ASCII launcher — no Chinese in the shell at all
open('launch.py', 'w').write(
    'import base64,os\\n'
    'open(\"gen_report.py\",\"wb\").write(base64.b64decode(\"' + payload + '\"))\\n'
    'os.system(\"python gen_report.py\")\\n'
)
"
python launch.py

Why this works: base64.b64decode(...) is pure ASCII in the shell. The Chinese characters are encoded as \uXXXX Unicode escapes (ASCII-safe) inside the script string, and Python’s textwrap.dedent + UTF-8 encode/decode reconstruct the original Chinese bytes correctly — the console encoding never touches the content.

Rule: On Windows, never put Chinese string literals inside a shell command (python -c "...", PowerShell heredoc, or echo). Always use base64 or Unicode escapes (\uXXXX) to carry non-ASCII content through the shell layer.

Subscripts and Superscripts

IMPORTANT: Never use Unicode subscript/superscript characters (₀₁₂₃₄₅₆₇₈₉, ⁰¹²³⁴⁵⁶⁷⁸⁹) in ReportLab PDFs. The built-in fonts do not include these glyphs, causing them to render as solid black boxes. Instead, use ReportLab’s XML markup tags in Paragraph objects:

import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "scripts"))
from setup_chinese_pdf import setup_chinese_pdf

from reportlab.platypus import Paragraph

# MUST be the very first reportlab operation
cn_font, styles = setup_chinese_pdf()

# Subscripts: use <sub> tag
chemical = Paragraph("H<sub>2</sub>O", styles['Normal'])

# Superscripts: use <super> tag
squared = Paragraph("x<super>2</super> + y<super>2</super>", styles['Normal'])

For canvas-drawn text (not Paragraph objects), manually adjust font the size and position rather than using Unicode subscripts/superscripts.

Command-Line Tools

pdftotext (poppler-utils)

# Extract text
pdftotext input.pdf output.txt

# Extract text preserving layout
pdftotext -layout input.pdf output.txt

# Extract specific pages
pdftotext -f 1 -l 5 input.pdf output.txt  # Pages 1-5

qpdf

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

# Split pages
qpdf input.pdf --pages . 1-5 -- pages1-5.pdf
qpdf input.pdf --pages . 6-10 -- pages6-10.pdf

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

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

pdftk (if available)

# Merge
pdftk file1.pdf file2.pdf cat output merged.pdf

# Split
pdftk input.pdf burst

# Rotate
pdftk input.pdf rotate 1east output rotated.pdf

Common Tasks

Extract Text from Scanned PDFs

# 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)

Add Watermark

from pypdf import PdfReader, PdfWriter

# Create watermark (or load existing)
watermark = PdfReader("watermark.pdf").pages[0]

# Apply to all pages
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)

Extract Images

# Using pdfimages (poppler-utils)
pdfimages -j input.pdf output_prefix

# This extracts all images as output_prefix-000.jpg, output_prefix-001.jpg, etc.

Password Protection

from pypdf import PdfReader, PdfWriter

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

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

# Add password
writer.encrypt("userpassword", "ownerpassword")

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

Quick Reference

Task	Best Tool	Command/Code
Merge PDFs	pypdf	`writer.add_page(page)`
Split PDFs	pypdf	One page per file
Extract text	pdfplumber	`page.extract_text()`
Extract tables	pdfplumber	`page.extract_tables()`
Create PDFs	reportlab	Canvas or Platypus
Command line merge	qpdf	`qpdf --empty --pages ...`
OCR scanned PDFs	pytesseract	Convert to image first
Fill PDF forms	pdf-lib or pypdf (see FORMS.md)	See FORMS.md

Next Steps

For advanced pypdfium2 usage, see REFERENCE.md
For JavaScript libraries (pdf-lib), see REFERENCE.md
If you need to fill out a PDF form, follow the instructions in FORMS.md
For troubleshooting guides, see REFERENCE.md

​PDF Processing Guide

​Overview

​Quick Start

​Python Libraries

​pypdf - Basic Operations

​Merge PDFs

​Split PDF

​Extract Metadata

​Rotate Pages

​pdfplumber - Text and Table Extraction

​Extract Text with Layout

​Extract Tables

​Advanced Table Extraction

​reportlab - Create PDFs

​Basic PDF Creation

​Create PDF with Multiple Pages

​Chinese / CJK PDF Content

The complete solution — setup_chinese_pdf()

Using setup_chinese_pdf() — Platypus (Paragraph / Table)

Using setup_chinese_pdf() — Canvas API