receipt-scanner skill

Overview

This skill extracts structured data from receipt images using OCR and pattern matching, returning vendor, date, line items, amounts, taxes, and totals in JSON or CSV. It supports common image and PDF formats and can run on single files or batches for bulk processing. The implementation focuses on reliable parsing with tested regex patterns and clean, exportable outputs.

How this skill works

The scanner runs OCR on each receipt image to get raw text, then applies configurable pattern matching and heuristics to identify vendor, transaction date, line items (description, quantity, unit price), taxes, and total amounts. It normalizes amounts and dates, validates numeric fields, and outputs a structured JSON object or tabular CSV. Batch mode iterates files and aggregates results into a single output file.

When to use it

• Digitize paper receipts for bookkeeping or expense reports
• Automate expense ingestion for accounting systems
• Extract line-level details for analytics or reimbursement
• Process large receipt collections in batch for audits or record keeping
• Generate structured datasets from receipts for ML training

Best practices

• Ensure images are clear, well-lit, and cropped to the receipt area for best OCR accuracy
• Prefer 300 DPI scans or high-resolution photos; avoid heavy glare and blur
• Provide consistent filename conventions to help trace original images in output
• Validate parsed totals against OCR numeric confidence for critical workflows
• Post-process with a small rules set for known vendors to increase accuracy

Example use cases

• Single receipt extraction to JSON for quick expense submission
• Batch convert a folder of receipts to a single CSV for accounting import
• Extract itemized lines to calculate category-level spend and tax
• Feed parsed receipts into an expense management API or database
• Create labeled datasets of receipt text and structured fields for model training

FAQ