When the security tool is part of the blast radius

The usual plan is detect more and patch faster.

That fails when the break includes the tool itself a bad path, a component with too much power, an update you don’t fully trust mid-incident.

Patching is recovery. Containment is what limits spread while something is already wrong.

I’ve been building host security in Utah with that framing: narrow actions, limits on the server, approval before bigger steps. Assume a part can fail; don’t let it own the rest of the box for free.

How do other Utah teams think about blast radius when the failure might include their own tooling?