LLM Hacker News 3d ago 2 min read
Agent Safehouse is an open-source macOS hardening layer that uses sandbox-exec to confine local coding agents to explicitly approved paths instead of inheriting a developer account’s full access.
Agent Safehouse is an open-source macOS hardening layer that uses sandbox-exec to confine local coding agents to explicitly approved paths instead of inheriting a developer account’s full access.
A popular Hacker News post highlighted Agent Safehouse, a macOS tool that wraps Claude Code, Codex and similar agents in a deny-first sandbox using sandbox-exec. The project grants project-scoped access by default, blocks sensitive paths at the kernel layer, and ships as a single Bash script under Apache 2.0.
Anthropic's Claude Code Cowork (multi-agent collaboration) feature was found to create a ~10GB VM bundle on macOS using Apple's Virtualization Framework without warning users. The GitHub issue garnered 200+ points on Hacker News.