Is this command safe to run?

Paste a shell command, SQL statement, or infra operation. You get its blast radius — how much it can destroy, whether it's reversible, and exactly which dangerous capability it triggers. The same classifier that powers infraveil-guard, the open-source seatbelt for AI agents.

Blast radius, not vibes

The commands that end companies aren't exotic — they're rm -rf in the wrong directory, a DELETE that forgot its WHERE, a force-push over shared history, a terraform destroy against prod. What they share is a large blast radius and no undo. This tool reads your command against a small, readable set of rules and tells you which of those it triggers and whether you can take it back.

It matters most for one reason: AI agents now run commands. The viral incidents — an agent that deleted a production database in seconds, backups and all — happened because nothing sat between “the model decided to run this” and “it ran.” The fix isn't to stop using agents; it's to make exactly these high-blast, irreversible actions wait for a human. That gate is the whole idea behind infraveil-guard and the Infraveil control plane.