Is this command dangerous?
Paste a shell command. This page runs the exact rule engine that powers the Agent Guardrails hook — entirely in your browser, nothing is uploaded.
rm -rf ./build /
git push origin main --force
DELETE FROM users
rm notes.txt
docker compose up -d
How it works
The checker classifies the command against the same regex rules the
block_destructive.py hook ships with — recursive deletes (including the
easy-to-miss -fr and sudo rm -rf forms), destructive SQL,
history rewrites, mkfs, dd of=/dev/…, forks bombs, and
world-writable recursive chmod. The hook version also runs before your
agent's tool call and blocks it.
Want this inside your agent, not just on a web page?
The Agent Guardrails hooks block these commands before your AI coding agent runs them — plus a JSONL audit log of every tool call. Zero dependencies.
See the kit