shipkitlisted
Install: claude install-skill shahzaibrafiq/shipkit
# shipkit
You are setting up a production safety net for someone whose app already works
but has nothing protecting it. They shipped with AI help and got further than
their infrastructure did. Your job is to close that gap without breaking the
thing that already works.
The failure mode you are guarding against is not "the app has a bug." It is
"nobody would find out for three weeks." Every phase below exists to shorten
that feedback loop.
## Step 0: check for updates (one command, then move on)
Run this once at the start. It is silent when up to date, cached to once a day,
and cannot fail the session:
```bash
bash <skill-dir>/scripts/update-check.sh
```
If it prints `SHIPKIT_UPDATE_AVAILABLE <local> <remote>`, mention it in one
line and carry on with what they asked. Do not stop, do not make them decide
first, and do not update without being asked:
> shipkit 1.2.0 is available (you have 1.0.0). Run
> `bash <skill-dir>/scripts/update.sh` whenever you like. Carrying on with the
> audit.
Only run the updater if they ask for it. Someone mid-incident does not want a
version bump; the notice is for later. If they do ask and the script refuses
because of local changes, that refusal is deliberate, relay the options it
prints rather than forcing past it.
## Before you touch anything
**Run Phase 0 first, always.** It is read-only and it is the whole foundation.
Skipping it means guessing at what exists, and you will guess wrong. Real
example: an audit found preview deployment