CSV reconciliation script spec
You need an AI agent to build a Python script that reconciles two CSV exports — a bank statement and your internal ledger — and emits a third CSV of unmatched transactions. The matching key is (date, amount, last-4-of-description), but real data has trailing whitespace, mixed date formats, and occasional duplicate rows. Write the prompt or spec you'd give the agent so it produces a correct, re-runnable script the first time. What would a vague one-line prompt ("write a script to reconcile these two CSVs") get wrong here?
reconcile_unmatched_rows(bank_rows: list[str], ledger_rows: list[str]) → list[str][["2026-01-31 | -125.00 | ACME CORP INV 1234 ","2026-02-01|-40.00|COFFEE 9981"],["01/31/2026|125.00|acme inv 1234","2026-02-03|17.25|BOOKS 4410"]]out["BANK|2026-02-01|40.00|9981","LEDGER|2026-02-03|17.25|4410"][["2026-03-01|10.00|SUB 7777","2026-03-01|10.00|SUB 7777"],["2026-03-01|10.00|sub 7777"]]out["BANK|2026-03-01|10.00|7777"][[" 2026-04-05 | $1,200 | VENDOR 8899 "],["4/5/2026|1200.00|vendor-8899"]]out[]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.