How-to
How to roll back a bad change
If a deploy goes bad, CatWrangler rolls it back automatically. You can also restore an earlier tagged state. And because every change must build, pass tests, and cover its decision before it merges, most broken code never reaches your trunk at all.
Bad changes happen. What matters is how little they cost you. CatWrangler treats rollback as a normal, fast operation, not a fire drill.
Recovery is built in, not bolted on
Most version-control horror stories start the same way: something broke, and getting back to a known-good state was harder than the original work. CatWrangler closes that gap by making the safe path the default one.
Validation comes first. A change can't reach the shared trunk unless it builds, passes its tests, and actually covers the decision it claims to make. That means a large share of bad changes are stopped before they're ever a problem you have to undo.
- A bad deploy rolls back automatically.
- An earlier tagged state can be restored when you want to step back further.
- Validation keeps broken code out of trunk, so most rollbacks never need to happen.
- 01
Let validation do the first catch
Before anything merges, the change must build, pass its tests, and cover the decision it claims. Code that can't clear that bar never reaches your trunk, so the most common bad changes are stopped before they ship.
- 02
Ship the change
On merge, your work builds, validates, and deploys itself. The new version swaps in with no downtime, so a release in progress never takes the running app offline.
- 03
Let a bad deploy roll back on its own
If a deploy goes bad, it rolls back automatically. You don't scramble to assemble a recovery, and there's no separate pipeline to babysit while it happens.
- 04
Restore a known-good tagged state
When you want to go back further than the last deploy, restore an earlier tagged state and it swaps back in.
- 05
Keep the why intact through the whole thing
Every change still carries the decision behind it, so after a rollback you and your agents can see exactly what was reverted and why, instead of guessing from a diff.
Related questions
Do I have to undo a bad change myself?
Usually no. A bad deploy rolls back automatically, and validation stops most broken code before it ever merges. When you want to step back further, you restore an earlier tagged state yourself.
Will rolling back take the app down?
Deploys swap in with no downtime. A bad deploy rolls back automatically, and you can restore an earlier tagged state when you want to go back deliberately.
Keep reading
Vibe-Engineering
Many agents. One codebase. Zero collisions.
Point your agents at CatWrangler and build — the discipline runs underneath.