"Safe by default with shell functions" documentation seems wrong?

[kevinrenskers]

The website has this text:

Safe by default with shell functions
Add these to your shell config and every agent runs inside Safehouse automatically — you don't have to remember. To run without the sandbox, use command claude to bypass the function.

And then has this code:

# ~/.config/fish/config.fish
function safe
    safehouse --add-dirs-ro="$HOME/mywork" $argv
end

# Sandboxed helpers without overriding the original binary names.
function sandbox-claude
    safe claude --dangerously-skip-permissions $argv
end

[...]

I am not sure in what way this fish config makes Claude "safe by default", or how it makes "every agent runs inside Safehouse automatically", without having to remember? You now have to run sandbox-claude instead of safehouse claude, if I'm not mistaken. But claude by itself isn't safe.

It seems the function name should shadow the agent name.

[davidfstr]

Indeed I think this documentation can be tightened to be more realistic.

I personally define a sclaude and scopilot command to run safehouse-sandboxed versions of Claude Code and Copilot CLI respectively. Typing 1 extra character is easy and more realisic (I agree) than typing sandbox-claude.