Lots of access, on purpose. And how to rein it in.
Claude Code starts in auto mode. More on that in a moment.
Claude asks. Codex builds a wall: its sandbox.
It sends what it reads. Files it never opens stay home.
A throwaway folder, 20 October: Claude Code 2.1.302, Codex 0.162.
Everything else in the test went ahead: edits, scripts, web pages, even deleting a folder git could restore.
/permissions, look, then Esc.Done early? “Show me every settings file you read, and explain each line.”
My setting: auto, everywhere.
.claude/settings.json, in the project folder:
You don’t write this. You ask:
Add a rule to this project’s settings so you can never read the interview transcripts.
In founders-study, from the git exercise:
/exit, start the agent again, and ask: “Read one of the interview transcripts.”Codex: ask the same. It may set up a permissions profile, or say it can’t.
Done early? Try to get round your own rule: “Search every file in this folder for the word pivot.” Did it read the interviews?
A rule stops the direct ways in, not every way. Real protection: keep it out of the folder.
claude.ai/settings/data-privacy-controlsLook at it. Decide. One minute.
Your data outside the folder. Bold where it’s cheap to undo.