Bank Statement Reconciler

The problem

Manual bookkeeping often leads to double-counting transfers as income, duplicate entries from overlapping CSV exports, and inconsistent categorization that ruins financial reports.

What it does

• Ingests raw CSV exports from Monzo, Starling, Revolut, Wise, and other banks with automatic format detection.
• Prevents double-counting by matching internal transfers (±3 days) and assigning unique transfer IDs.
• Enforces strict deduplication using a uniqueness key of account, date, amount, and normalized description.
• Generates monthly close reports with net totals, category deltas, and anomaly detection for large transactions.
• Maintains a persistent rules.csv to automate categorization based on your historical decisions.

Why this beats prompting it yourself

General LLM prompts often hallucinate balances or fail to handle the complex logic of multi-leg transfers across different accounts. This skill implements a rigorous append-and-correct ledger system that ensures data integrity and produces figures an accountant can actually use.

Use cases

• Consolidating multiple bank exports into a single source of truth ledger.
• Preparing clean income and spending reports for tax returns or accounting.
• Auditing personal or business spending to find subscription leaks and anomalies.
• Splitting business and personal expenses within a single shared bank account.

Known limitations

Requires transaction data to be provided in CSV format. Date ambiguity in files (e.g., 01/02/2024) may require a one-time user confirmation if the file lacks clear day-month indicators.