What problem does it solve? Adding a new bank, wallet, or platform bill export to finance-tracker often causes mis-booked accounts, missed refund pairs, and false transfer matches. This Skill provides a proven workflow for onboarding statement sources so rows import as correct formal facts and participate correctly in relation scanning. ## Core Features & Use Cases - Taxonomy bucket method: Partition real bills by source-native fields (e.g. Alipay status × direction, WeChat status × type) into a bucket table before writing any scan rules, avoiding global keyword soup. - Account semantics guardrails: Fix which account each transaction books to (withdrawals, money-market redemptions, credit repayments) before tuning scan rules. - Phase-scoped scan rules: Assign hard-key refunds to Phase A, payment mirrors to Phase B, transfers to Phase C, and bank refunds to Phase D, with bucket-scoped predicates. - Use Case: When adding a new bank's CSV export, build the bucket table from real multi-year statements, define per-bucket import transforms and scan roles, then calibrate with a fresh SQLite database and relations check until pending counts stabilize. ## Quick Start Ask the assistant to onboard a new bank statement export for finance-tracker using the statement source onboarding workflow, starting from real sample bills.
