Show HN: Safer – Sleep better while AI agents have shell access

Category: security

Tags: ai-agents, security, cli-tool, go, guardrails

Score: 6.8/10 (Innovation: 7, Technical: 5, Documentation: 8, Utility: 7)

Safer is a command wrapper for AI coding agents that blocks or warns before executing risky shell commands like file deletion, database mutations, or infrastructure changes. It introduces a capability-based pre-execution guardrail system that is novel in the AI agent tooling space, combining read-only defaults with explicit permission flags to help developers sleep better while agents have shell access.

Target audience: backend devs, devops, AI/ML engineers

Repository: https://github.com/crufter/safer · Go · MIT · 7 stars

View on Hacker News