Show HN: GuardRail – 13 guards that stop Claude Code before it pushes to main

Category: security

Tags: ai-agents, claude-code, safety-guards, shell-hooks, devtools

Score: 6.0/10 (Innovation: 6, Technical: 5, Documentation: 7, Utility: 6)

GuardRail is a set of bash-based pre-execution hooks for Claude Code that block dangerous shell commands (force pushes to main, rm -rf on protected paths, DELETE without WHERE, secret exfiltration) before they run, and log every block to an audit trail. It targets an emerging real gap — agent safety at the execution layer rather than at the prompt/response layer — and ships 13 free MIT guards with a one-command installer. The project is early (0 stars, single HN vote) and partly gated behind a Pro tier, but the concept is timely and the failure modes it addresses are concrete.

Target audience: backend devs, devops, platform engineers using AI coding agents

Repository: https://github.com/FvdHMBAI/guardrail · Shell · MIT

View on Hacker News