Automation that breaks quietly is worse than none. ZwayRank's crew includes a debugger, a doctor and a CTO, and it tells you plainly when something needs you.
Alerts: every problem with what happened, why, what the crew did about it, and whether it needs you, with a Fix button where one applies.
Crew health: crashes, muted tasks, failed steps, and the nightly self-test result.
The doctor's log: each remedy it applied and whether it worked.
Emails: only for the things the crew could not fix, capped so your inbox stays sane.
A publish fails because a host was slow. A scan fails because a model returned nothing. A key expires. Left alone, each of those becomes a silent gap in the work. The doctor runs on every pass, recognises the known failures and applies the remedy: retry, reconnect, clear the stuck job, close the alert that was already solved.
Some failures are new. Those go to Forge, the CTO agent, with the trace and the context. It reads the code that failed, writes up the cause and the fix, and the owner's team ships it. You see it in the record as a note, not as a broken screen.
Every night the self-test calls every read endpoint for every brand and checks the invariants: the database is healthy, backups are fresh, no job is stuck, no text carries a forbidden character. What it finds becomes suggestions for the doctor and, if needed, an alert for you.
An alert says what happened, why, what was done, and what, if anything, needs you. The ones that need nobody close themselves with the reason. Only needs-you alerts and serious failures reach your inbox, at most a dozen a day.
The third failure of the same task in a day raises one alert and one email, and the CTO agent gets the trace.
Add one brand. The first scan is free and takes a few minutes. Then you decide whether the crew gets to work.
Seven days. One brand. Your keys after the first scan.
ZwayRank