The original title is: "gate.cat: a deterministic, fail-closed veto that stops AI coding agents before rm -rf"

Short summary

gate.cat is an open-source, deterministic, fail-closed veto tool that blocks irreversible shell commands (rm -rf, DROP TABLE, terraform destroy, secret exfiltration) before AI coding agents execute them. It uses string and path analysis with no model call in the veto path, making it immune to prompt injection. Tested against 1M+ real agent commands with zero misses, it integrates with Claude Code hooks, any CLI agent via shell wrapper, or any OpenAI-compatible API via proxy.

• •Deterministic, fail-closed gate that blocks dangerous shell commands before execution — no LLM in the veto path
• •Tested on 1M+ real agent commands with 0 misses; publishes its own known gaps for transparency
• •Integrates with Claude Code hooks, CLI agents via shell wrapper, and OpenAI-compatible APIs via local proxy

Generated with AI, which can make mistakes.