Set up data cleanup jobs for my app. Inventory the tables or collections that grow unbounded: sessions, password reset tokens, email verification tokens, magic link tokens, logs, webhook event records, abandoned carts, expired invites, soft-deleted rows. For each, add a scheduled cleanup that deletes rows past a sensible retention window: sessions and tokens expire within hours or days, logs keep weeks, soft deletes purge after a grace period. Run them in small batches with LIMIT so the cleanup itself doesn't lock tables, and schedule them at low-traffic times. Log what each run deleted so the job is auditable. Don't delete anything tied to money or legal obligations: payment records, invoices, and audit trails stay. If a record type has regulatory retention requirements in my domain, flag it instead of guessing. Use whatever scheduling mechanism already exists in the stack (cron, platform scheduled functions, or a worker); don't add a new service. Deliver: the retention policy per data type with reasoning, the cleanup jobs and schedule, and verification: insert expired test rows, run a cleanup, confirm they're gone and recent rows survive.