← All projects
Automations That Run on a Scheduleadvanced · ~110 min

Forget On Schedule — a Nightly Retention Sweeper You Can Actually Prove

Every other scheduled automation you build collects, watches, copies, or sends — and if it misbehaves, you rerun it. What happens when the scheduled job's whole purpose is to *remove* things?

Keeping everything forever is a decision too, and it is the one nobody makes on purpose: old candidate CVs, scanned IDs, exports, and backups pile up until you cannot answer a simple question like "why do you still have this?" This project is about building the one kind of automation where a bug cannot be undone by running it again — so it teaches the shape that makes deletion survivable: a schedule written as data, a grace period before anything is truly gone, holds that outrank the schedule, a hard cap that stops a bad rule from emptying the room in one night, and a paper trail that lets you explain any single decision six months later.

🤝 Want this built for you? Start a request — our AI scopes it, estimates a price range, and matches you with a vetted AI Advisor.
Share this:
Forget On Schedule — a Nightly Retention Sweeper You Can Actually Prove
Unlock this walkthrough

Two ways to open it

Just this build
$10 one-time

Unlocks this full walkthrough on your account — no subscription, and no account needed until you check out.

Best value Over 150 projects + the full course
$99 / year

Over 150 project walkthroughs, the 6-week course, and your certificate.

🔒 Secure checkout by Stripe · Already a member? Sign in

What you'll do

A nightly retention sweeper you own end to end: a plain-text retention schedule that states, per class of data, which date starts its clock and how long it lives; a read-only plan mode that is the default and shows you every candidate before anything moves; guardrails that refuse to leave your declared roots, refuse to follow symlinks, refuse to cross filesystems, and stop the run cold when a rule would act on more than your cap allows; a two-phase expiry where items first move atomically into a dated waiting area and only become permanently gone after a grace window and a verified-backup gate; holds that outrank every rule and are re-checked in the last instant before the irreversible step; an append-only JSONL ledger plus an explain command that reconstructs any single decision months later; a restore command you have actually used; and three drills you have actually run — a runaway rule stopped by the cap, a sweeper killed mid-apply and repaired by reconcile, and a held item that survived its own due date.