import

CSV and OFX transaction import parsing. Use when working on files in src/lib/import/.

⭐2stars🔀0forks📁View on GitHub🕐Updated Jan 11, 2026⬇️1downloads

When & Why to Use This Skill

This Claude skill streamlines the complex process of parsing and normalizing financial transaction data from CSV and OFX files. It automates critical financial workflows including currency conversion to minor units, intelligent duplicate detection using fuzzy matching, and robust data validation to ensure high integrity for accounting and fintech applications.

Use Cases

  • Automated Bank Statement Integration: Efficiently parse and standardize transaction records from various bank exports (CSV/OFX) into a unified system format.
  • Financial Data Normalization: Automatically handle diverse date/number formats and convert currency amounts into precise minor units (e.g., cents) for consistent database storage.
  • Intelligent Duplicate Prevention: Identify and filter redundant transactions during import using Levenshtein distance logic based on date windows, amount tolerance, and description similarity.
  • Robust Data Cleaning and Validation: Resolve character encoding issues (UTF-8, Latin-1) and validate parsed data against strict financial rules to prevent corrupted entries.
nameimport
descriptionCSV and OFX transaction import parsing. Use when working on files in src/lib/import/.

Import Guidelines

Flow

File Drop → Parse (CSV/OFX) → Column Mapping → Formatting → Preview → Import

Money Handling

All amounts stored as integers in minor units:

import { toMinorUnitsForCurrency } from "@/lib/domain/currency";

toMinorUnitsForCurrency(12.34, "USD"); // 1234 (cents)
toMinorUnitsForCurrency(1234, "JPY");  // 1234 (0 decimals)
toMinorUnitsForCurrency(1.234, "KWD"); // 1234 (3 decimals)

Currency resolution order: OFX CURDEF → Account → User default → Vault default → USD

CSV Parser

Auto-detects: separator, headers, date/number formats.

Duplicate Detection

Uses Levenshtein distance: date window ±3 days, amount tolerance 1 cent, description 80% similarity.

Critical Rules

  • Never trust user input - validate all parsed data
  • Handle encoding: UTF-8, Latin-1, Windows-1252
  • Preserve originalRow for debugging
  • Handle negative formats like (123.45)
  • Return structured errors, don't throw

Testing

Test with real bank exports covering different date formats (US/EU/ISO), number formats, edge cases, large files.