bank-statement-processing

What problem does it solve?

This Skill removes the manual effort of downloading Wells Fargo statement PDFs, extracting transactions, categorizing them, and loading the results into SerenDB for later analysis.

Core Features & Use Cases

• Runtime-authenticated PDF retrieval: prompts for Wells Fargo login details (including OTP/passkey/manual handoff) and downloads statements as local PDFs using SerenDesktop Local Playwright MCP.
• Transaction parsing + normalization: extracts statement metadata and converts statement rows into structured transactions with dates, amounts, and normalized descriptions.
• Rules-first categorization with optional LLM fallback: applies keyword rules first, then uses heuristic LLM classification, with support for an optional remote LLM endpoint.
• Masked SerenDB persistence: writes masked account fields and structured statement/transaction data into SerenDB read-only tables and generates reports/JSONL exports for downstream reuse.

Quick Start

Run it to download and process your last 12 months of Wells Fargo statements with: python3 scripts/run.py --config config.json --mode read-only --months 12 --out artifacts/wellsfargo