invoice-processing

name	invoice-processing
description	OCR and extract invoice/receipt data → validate → output organized spreadsheet. Use when processing invoices, receipts, or expense documents.

Invoice Processing

Source: $ARGUMENTS (defaults: ./inbox/ → ./output/invoices.xlsx)

Deterministic pipeline: scan documents → extract fields → validate → write spreadsheet.

Workflow

1. Discover documents — glob $SOURCE_DIR for *.pdf, *.png, *.jpg, *.jpeg
2. Extract fields — for each document read/OCR and extract:
   • vendor (company name)
   • date (normalize to YYYY-MM-DD)
   • amount (numeric, preserve currency symbol)
   • invoice_number (as printed; null if absent)
   • category (auto-classify: travel, office, software, meals, other)
3. Validate — flag anomalies before writing:
   • Duplicate invoice_number across documents
   • amount = 0 or negative
   • date more than 90 days in the past or any future date
   • Missing required fields (vendor, date, amount)
4. Write spreadsheet — create $OUTPUT_FILE with columns: File | Vendor | Date | Amount | Currency | Invoice# | Category | Flag
   • Sort by Date ascending
   • Highlight flagged rows (add [FLAG: reason] in Flag column)
5. Print summary — total count, total amount per currency, flagged count

Input Contract

$SOURCE_DIR/
  *.pdf | *.png | *.jpg | *.jpeg   — invoice/receipt files

Output Contract

$OUTPUT_FILE.xlsx    — populated spreadsheet, sorted by date
stdout               — summary: N docs processed, N flagged, totals by currency

Constraints

• Do NOT move or delete source files
• Do NOT push to accounting system — output only (human reviews before upload)
• If a document cannot be read, add row with [UNREADABLE] flag and continue
• Currency: preserve original; do not convert