Legacy payment module migration
You're tasked with migrating a 4,000-line untested legacy Java payment-processing module to a new internal API, and you want to lean on an AI agent for the bulk of it. How do you decompose this so the agent is useful without it confidently breaking money-handling logic? What do you delegate, what do you keep, and where are the checkpoints?
Implement
plan_migration_steps(slice_lines: list[str]) → list[str]Examples
in
[["rounding_util|yes|120|no","csv_export|no|300|no","status_machine|yes|450|no"]]out["characterize:rounding_util","characterize:csv_export","characterize:status_machine","migrate_human:rounding_util","migrate_agent:csv_export","migrate_human:status_machine","shadow_parity","cutover"]in
[["logging|no|80|yes","fee_math|yes|60|no"]]out["characterize:fee_math","migrate_human:fee_math","migrate_agent:logging","shadow_parity","cutover"]What a strong answer looks like
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.
0:00 of about 24 min
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.