Rust Tool 'destructive_command_guard' Blocks Dangerous Shell Commands Run by AI Agents

A developer has released a Rust-based tool called destructive_command_guard, designed to act as a safety layer between AI coding agents and shell access. The tool intercepts potentially catastrophic commands — such as rm -rf, git push --force, DROP TABLE, and terraform destroy — using regex pattern matching before they reach the shell. Its development was partly inspired by a real incident in which an AI agent belonging to entrepreneur Matt Shumer wiped days of files and code due to a single incorrect path variable. Unlike permission systems or audit logs alone, the guard returns a structured error to the agent when a command is blocked, allowing it to log the event or escalate to a human. The project addresses a growing concern that AI agents with unrestricted shell access can cause irreversible damage even when they have operated safely hundreds of times before.