What you are trying to do
You want the work to keep moving without you watching it. You also want to know, by Monday morning, what happened over the weekend, what got stuck, and what you should look at first when you sit down with coffee.
Routines are the recurring jobs Ship runs in the background to make that true. There are three of them you should know by name.
Self-heal — every hour, fix one stuck thing
Pipelines stall in small, boring ways. A ticket gets stuck because nobody set a label. A pull request sits open with green tests for three days because nobody noticed. A repository's scheduled job got silently disabled because it failed fifty times in a row.
The self-heal routine runs every hour. Its only job is to scan for state that has stopped moving and unblock the smallest concrete thing it can. If it finds a ticket sitting in progress for a day with no activity and a missing state label, it sets the label, leaves a one-line comment explaining why, and stops. If it finds a workflow that has been silently disabled, it does not turn it back on — it writes you an Inbox letter naming the exact command to run.
The rule is one fix per run. Small and often, never big and rare. By the time you sit down on Tuesday, the seven small things that would have caused a fire have quietly been put back on the rails.
Daily digest — every morning, one letter in your Inbox
Once a day, the daily digest routine reads the last 24 hours of activity across every repository in the workspace and files one Inbox letter. The letter has five sections, always in the same order: a headline (green, yellow or red, with one sentence why), a throughput count (how many finishes, how many dispatches, how many things got stuck in flight), two to four lessons learned, any open pull requests or blocked tickets that have been sitting too long, and how many Inbox letters from prior digests you still have not read.
This is your morning newsletter. It is intentionally one item, not five. If you have time for nothing else, the daily digest is the thing to open.
The retro will still file on a silent day — a one-line "all green, nothing to report" letter — because silence is ambiguous. You should be able to glance at your Inbox and know that the routine actually ran.
Weekly audit — every Monday, the assurance pass
Delivery work and the work of proving delivery happened well are not the same job. The weekly audit routine runs once a week and produces a separate report covering four areas: tech-debt and architecture drift, QA coverage gaps, security findings, and process health.
It does two things at once. It files an audit report in your Inbox with the headline numbers — how many gaps, where the worst ones are, which pull requests have been sitting open the longest. And it opens tickets for the gaps worth tracking — usually fewer than ten in a week, capped on purpose so the board does not drown.
The point of the weekly audit is to keep "we should clean that up sometime" from being the most expensive sentence in your codebase. By the time Monday morning arrives, the cleanup work is already a real ticket sitting in your backlog with a body that names the file and the reason.
What this changes about your week
You stop being the clock. You stop being the one who notices on Thursday that an important pull request has been sitting unmerged since Monday.
Every morning you open the Inbox and there is a small, predictable handful of letters: yesterday's digest, anything self-heal could not auto-fix, anything an agent escalated overnight. By the time you finish your coffee you have processed the whole list. Routines are how Ship breathes while you are not watching — you did not hire an operations team for this, you bought one.