Extract PDF Transactions

Extract PDF Transactions Skill

Extract transaction data from PDF bank and credit card statements.

Skill Metadata

• Name: extract-pdf-transactions
• Category: Financial Data Processing
• Complexity: Medium
• Privacy: Local processing only, no external APIs

Capabilities

This skill extracts structured transaction data from PDF financial statements:

1. Multi-format PDF parsing - Supports various bank statement layouts
2. Table detection - Automatically identifies transaction tables
3. Data extraction - Parses dates, descriptions, amounts, and balances
4. Multi-page support - Processes statements spanning multiple pages
5. Error handling - Gracefully handles malformed PDFs
6. Validation - Verifies extracted data quality

Usage

/finance-process extract --input="~/Documents/Finance/*.pdf" --output="~/Documents/Finance/transactions.csv"

How It Works

1. PDF Discovery: Finds all PDF files matching the input pattern
2. Content Extraction: Uses pdfplumber to extract text and tables
3. Table Parsing: Identifies transaction tables using pattern matching
4. Data Normalization: Standardizes column names and formats
5. Validation: Checks for completeness and accuracy
6. CSV Export: Saves transactions to specified output file

Script

The main script is scripts/extract_pdf_statements.py which:

• Accepts PDF file paths (glob patterns supported)
• Extracts transaction tables
• Outputs to CSV format
• Provides progress feedback and error reporting

Dependencies

Required Python packages:

• pdfplumber (primary extraction engine)
• pandas (data manipulation)
• python-dateutil (date parsing)
• tabula-py (fallback for complex tables)

Output Format

The skill produces CSV files with these columns:

• date - Transaction date (YYYY-MM-DD)
• description - Transaction description/merchant
• amount - Transaction amount (negative for debits)
• balance - Account balance after transaction
• category - Auto-categorized transaction type
• source_file - Original PDF filename

Error Handling

• Logs files that couldn't be processed
• Reports extraction quality metrics
• Provides suggestions for failed extractions
• Continues processing remaining files on error

Privacy & Security

All PDF processing happens locally:

• No cloud API calls
• No data transmission
• Files stay on local filesystem
• Sensitive data never leaves the machine