This is your go-to for pulling structured data out of PDFs without losing your mind. It wraps pdfplumber to extract text with character-level positioning, detect tables automatically, and let you crop, filter, and debug what's actually in the PDF. The visual debugging tools are genuinely helpful when table detection goes sideways. Works great for financial reports, invoices, and any PDF where you need the actual table data instead of mangled copy-paste text. Won't help with scanned PDFs though, those need OCR first. The included patterns for converting tables to DataFrames and filtering by font or region save you from rewriting the same boilerplate every time.
npx -y skills add claude-office-skills/skills --skill pdf-extraction --agent claude-codeInstalls into .claude/skills of the current project.
This skill enables precise extraction of text, tables, and metadata from PDF documents using pdfplumber - the go-to library for PDF data extraction. Unlike basic PDF readers, pdfplumber provides detailed character-level positioning, accurate table detection, and visual debugging.
Example prompts:
import pdfplumber
# Open PDF
with pdfplumber.open('document.pdf') as pdf:
# Access pages
first_page = pdf.pages[0]
# Document metadata
print(pdf.metadata)
# Number of pages
print(len(pdf.pages))
PDF Document
├── metadata (title, author, creation date)
├── pages[]
│ ├── chars (individual characters with position)
│ ├── words (grouped characters)
│ ├── lines (horizontal/vertical lines)
│ ├── rects (rectangles)
│ ├── curves (bezier curves)
│ └── images (embedded images)
└── outline (bookmarks/TOC)
with pdfplumber.open('document.pdf') as pdf:
# Single page
text = pdf.pages[0].extract_text()
# All pages
full_text = ''
for page in pdf.pages:
full_text += page.extract_text() or ''
# With layout preservation
text = page.extract_text(
x_tolerance=3, # Horizontal tolerance for grouping
y_tolerance=3, # Vertical tolerance
layout=True, # Preserve layout
x_density=7.25, # Chars per unit width
y_density=13 # Chars per unit height
)
# Extract words with positions
words = page.extract_words(
x_tolerance=3,
y_tolerance=3,
keep_blank_chars=False,
use_text_flow=False
)
# Each word includes: text, x0, top, x1, bottom, etc.
for word in words:
print(f"{word['text']} at ({word['x0']}, {word['top']})")
# Get all characters
chars = page.chars
for char in chars:
print(f"'{char['text']}' at ({char['x0']}, {char['top']})")
print(f" Font: {char['fontname']}, Size: {char['size']}")
with pdfplumber.open('report.pdf') as pdf:
page = pdf.pages[0]
# Extract all tables
tables = page.extract_tables()
for i, table in enumerate(tables):
print(f"Table {i+1}:")
for row in table:
print(row)
# Custom table detection
table_settings = {
"vertical_strategy": "lines", # or "text", "explicit"
"horizontal_strategy": "lines",
"explicit_vertical_lines": [], # Custom line positions
"explicit_horizontal_lines": [],
"snap_tolerance": 3,
"snap_x_tolerance": 3,
"snap_y_tolerance": 3,
"join_tolerance": 3,
"edge_min_length": 3,
"min_words_vertical": 3,
"min_words_horizontal": 1,
"intersection_tolerance": 3,
"text_tolerance": 3,
"text_x_tolerance": 3,
"text_y_tolerance": 3,
}
tables = page.extract_tables(table_settings)
# Find tables (without extracting)
table_finder = page.find_tables()
for table in table_finder:
print(f"Table at: {table.bbox}") # (x0, top, x1, bottom)
# Extract specific table
data = table.extract()
# Create visual debug image
im = page.to_image(resolution=150)
# Draw detected objects
im.draw_rects(page.chars) # Character bounding boxes
im.draw_rects(page.words) # Word bounding boxes
im.draw_lines(page.lines) # Lines
im.draw_rects(page.rects) # Rectangles
# Save debug image
im.save('debug.png')
# Debug tables
im.reset()
im.debug_tablefinder()
im.save('table_debug.png')
# Define bounding box (x0, top, x1, bottom)
bbox = (0, 0, 300, 200)
# Crop page
cropped = page.crop(bbox)
# Extract from cropped area
text = cropped.extract_text()
tables = cropped.extract_tables()
# Filter characters by region
def within_bbox(obj, bbox):
x0, top, x1, bottom = bbox
return (obj['x0'] >= x0 and obj['x1'] <= x1 and
obj['top'] >= top and obj['bottom'] <= bottom)
bbox = (100, 100, 400, 300)
filtered_chars = [c for c in page.chars if within_bbox(c, bbox)]
# Get text by font
def extract_by_font(page, font_name):
chars = [c for c in page.chars if font_name in c['fontname']]
return ''.join(c['text'] for c in chars)
# Extract bold text (often "Bold" in font name)
bold_text = extract_by_font(page, 'Bold')
# Extract by size
large_chars = [c for c in page.chars if c['size'] > 14]
with pdfplumber.open('document.pdf') as pdf:
# Document metadata
meta = pdf.metadata
print(f"Title: {meta.get('Title')}")
print(f"Author: {meta.get('Author')}")
print(f"Created: {meta.get('CreationDate')}")
# Page info
for i, page in enumerate(pdf.pages):
print(f"Page {i+1}: {page.width} x {page.height}")
print(f" Rotation: {page.rotation}")
to_image() to understand PDF structureimport pandas as pd
def pdf_tables_to_dataframes(pdf_path):
"""Extract all tables from PDF as pandas DataFrames."""
dfs = []
with pdfplumber.open(pdf_path) as pdf:
for i, page in enumerate(pdf.pages):
tables = page.extract_tables()
for j, table in enumerate(tables):
if table and len(table) > 1:
# First row as header
df = pd.DataFrame(table[1:], columns=table[0])
df['_page'] = i + 1
df['_table'] = j + 1
dfs.append(df)
return dfs
def extract_invoice_amount(pdf_path):
"""Extract amount from typical invoice layout."""
with pdfplumber.open(pdf_path) as pdf:
page = pdf.pages[0]
# Search for "Total" and get nearby numbers
words = page.extract_words()
for i, word in enumerate(words):
if 'total' in word['text'].lower():
# Look at next few words
for next_word in words[i+1:i+5]:
text = next_word['text'].replace(',', '').replace('$', '')
try:
return float(text)
except ValueError:
continue
return None
def extract_columns(page, num_columns=2):
"""Extract text from multi-column layout."""
width = page.width
col_width = width / num_columns
columns = []
for i in range(num_columns):
x0 = i * col_width
x1 = (i + 1) * col_width
cropped = page.crop((x0, 0, x1, page.height))
columns.append(cropped.extract_text())
return columns
import pdfplumber
import pandas as pd
def extract_financial_tables(pdf_path):
"""Extract tables from financial report and save to Excel."""
with pdfplumber.open(pdf_path) as pdf:
all_tables = []
for page_num, page in enumerate(pdf.pages):
# Debug: save table visualization
im = page.to_image()
im.debug_tablefinder()
im.save(f'debug_page_{page_num+1}.png')
# Extract tables
tables = page.extract_tables({
"vertical_strategy": "lines",
"horizontal_strategy": "lines",
"snap_tolerance": 5,
})
for table in tables:
if table and len(table) > 1:
# Clean data
clean_table = []
for row in table:
clean_row = [cell.strip() if cell else '' for cell in row]
clean_table.append(clean_row)
df = pd.DataFrame(clean_table[1:], columns=clean_table[0])
df['Source Page'] = page_num + 1
all_tables.append(df)
# Save to Excel with multiple sheets
with pd.ExcelWriter('extracted_tables.xlsx') as writer:
for i, df in enumerate(all_tables):
df.to_excel(writer, sheet_name=f'Table_{i+1}', index=False)
return all_tables
tables = extract_financial_tables('annual_report.pdf')
print(f"Extracted {len(tables)} tables")
import pdfplumber
import re
from datetime import datetime
def extract_invoice_data(pdf_path):
"""Extract structured data from invoice PDF."""
data = {
'invoice_number': None,
'date': None,
'total': None,
'line_items': []
}
with pdfplumber.open(pdf_path) as pdf:
page = pdf.pages[0]
text = page.extract_text()
# Extract invoice number
inv_match = re.search(r'Invoice\s*#?\s*:?\s*(\w+)', text, re.IGNORECASE)
if inv_match:
data['invoice_number'] = inv_match.group(1)
# Extract date
date_match = re.search(r'Date\s*:?\s*(\d{1,2}[/-]\d{1,2}[/-]\d{2,4})', text)
if date_match:
data['date'] = date_match.group(1)
# Extract total
total_match = re.search(r'Total\s*:?\s*\$?([\d,]+\.?\d*)', text, re.IGNORECASE)
if total_match:
data['total'] = float(total_match.group(1).replace(',', ''))
# Extract line items from table
tables = page.extract_tables()
for table in tables:
if table and any('description' in str(row).lower() for row in table[:2]):
# Found line items table
for row in table[1:]: # Skip header
if row and len(row) >= 3:
data['line_items'].append({
'description': row[0],
'quantity': row[1] if len(row) > 1 else None,
'amount': row[-1]
})
return data
invoice = extract_invoice_data('invoice.pdf')
print(f"Invoice #{invoice['invoice_number']}")
print(f"Total: ${invoice['total']}")
import pdfplumber
def parse_resume(pdf_path):
"""Extract structured sections from resume."""
with pdfplumber.open(pdf_path) as pdf:
full_text = ''
for page in pdf.pages:
full_text += (page.extract_text() or '') + '\n'
# Common resume sections
sections = {
'contact': '',
'summary': '',
'experience': '',
'education': '',
'skills': ''
}
# Split by common headers
import re
section_patterns = {
'summary': r'(summary|objective|profile)',
'experience': r'(experience|employment|work history)',
'education': r'(education|academic)',
'skills': r'(skills|competencies|technical)'
}
lines = full_text.split('\n')
current_section = 'contact'
for line in lines:
line_lower = line.lower().strip()
# Check if line is a section header
for section, pattern in section_patterns.items():
if re.match(pattern, line_lower):
current_section = section
break
sections[current_section] += line + '\n'
return sections
resume = parse_resume('resume.pdf')
print("Skills:", resume['skills'])
pip install pdfplumber
# For image debugging (optional)
pip install Pillow
larksuite/cli
googleworkspace/cli
googleworkspace/cli
googleworkspace/cli