[Feature Request] Improve sandbox security guidance for Python script execution environments

Bug Description
"I am planning to ship an application that can run python scripts. It ships it's own python interpreter to make sure I know what is being run. Fable flagged that a potential security issue is that I have not safeguarded the execution of user scripts that access of modify system files (e.g. .ssh) or access the internet - none of which my application requires. So it filed a bug report called sandbox-plan. It was now asked to review how I can implement good safeguards against harm for the user of the application. This flagged Fable 5's safeguards as unsafe. Unfortunately plugging sandbox related problems are a vital use of claude, so this is probably not a good place to flag."

Environment Info

• Platform: darwin
• Terminal: xterm-256color
• Version: 2.1.226
• Feedback ID: ef451d43-29b8-40fc-9c4d-5ca9964f6cb2

Errors

[]