Automations That Run on a Schedulemaster · ~240 min

Seven Jobs, One Door — a Nightly Pass That Runs In Order, Inside a Deadline, and Reports What It Refused to Run

You did the hard part already — you have a nightly backup, a retention sweeper, a page watcher, a weekly report, a morning digest. So why does the whole thing still feel like luck?

Because six separate scheduler entries are six programs that cannot see each other: the sweeper deletes on time whether or not the backup succeeded, the digest emails a total that was never recomputed, and a job that has been broken for nine days looks exactly like a job that had nothing to do. This project is the capstone that turns a pile of automations into one operated system — a single nightly pass that knows the order, refuses to run a job whose inputs are stale or missing, respects a deadline instead of spilling into your working day, benches what is broken so it stops burning the night, escalates by how much a job actually matters, and hands you one honest page every morning that lists what it refused to do and why. It ends with the test that separates a hobby from an operation: somebody else fixes it from your runbook while you say nothing.

🤝 Want this built for you? Start a request — our AI scopes it, estimates a price range, and matches you with a vetted AI Advisor. →

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 160 projects + the full course

$99 / year

Over 160 project walkthroughs, the 6-week course, and your certificate.

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

What you'll do

A single nightly orchestrator you built with Claude Code that owns every scheduled job you already have — one manifest declaring each job's command, owner, criticality, timeout, resources, and preconditions; a reviewable execution plan produced by a real topological sort that refuses to configure a cycle; three kinds of precondition (upstream succeeded, upstream succeeded recently enough, and a live check that must pass right now) so no job ever runs on stale or absent input; per-job timeouts that kill the whole process group rather than leaking orphans; a pass deadline with admission control that declines to start work it cannot finish and records the decision instead of overrunning into your morning; resource tags so two jobs that need the same drive never run together; automatic benching of a job that fails N passes in a row, with its downstream honestly blocked rather than silently proceeding; an escalation ladder that pages on transitions and severity rather than on every occurrence, and tells you when something clears; a one-page morning report whose first section is what did not run and why; one scheduler entry replacing all the old ones, verified by a one-door audit; a heartbeat the runner cannot fake; a `nightly why <job>` command that reconstructs any single decision; and a generated runbook that a second person has actually used to restore service without your help.