macOS Seatbelt sandbox CLI for developers. Protect credentials (SSH, AWS, GPG) from malicious npm packages, supply chain attacks, and untrusted build scripts. Deny-by-default filesystem isolation. Perfect for Claude Code agentic workflows with --dangerously-skip-permissions.

A Python decorator that executes untrusted code in isolated sandboxes with defense-in-depth security, designed for safe execution of LLM-generated code on local machines.

Fearlessly build arbitrary untrusted Cargo projects

Uses sandboxing feature of MacOS to sandbox tools/scripts

Sandboxes Minecraft on MacOS, protecting from malicious code.

lightweight OS-level sandbox for securing processes with filesystem and network controls - zero containers, zero fuss

PoC of sandboxing for the Antigravity AI IDE on OSX. PRs welcome to improve the sandboxing

Run Ralph (Claude, Codex, opencode, Cursor) on macOS using sandbox-exec instead of Docker.

Per-task kernel-layer sandbox for AI coding agents on macOS. Wraps any agent (Claude Code, Codex, Cursor, OpenClaw) so prompt-injection-driven credential exfiltration fails at the syscall layer, even when the agent has bypassPermissions.

Per-process OS-level sandboxes for Go: macOS sandbox-exec (SBPL/seatbelt) + Linux bubblewrap, applied to an *exec.Cmd via a declarative Profile. Non-optional SBPL literal validator; narrowed bwrap mounts + namespace unsharing.