Feature request: Add support for Claude Code's sandboxing

[edahlseng]

Anthropic recently released sandboxing for Claude Code, but it looks like the /sandbox command isn't supported in Zed yet. Adding support for this would be a huge benefit for security and productivity and would address some of the concerns from issues like #37343.

[elidickinson]

That seems like a good place to start since they open sourced an abstraction layer on top of the OS level tools (bubblewrap for linux and seatbelt/sandbox-exec for mac): https://github.com/anthropic-experimental/sandbox-runtime

Codex appears to have gone with a direct integration with LinuxSeccomp and Seatbelt. https://github.com/openai/codex/blob/79aa83ee399353c27485db85b6464d36246b3fcc/codex-rs/core/src/safety.rs