permission modes:
decide when it asks.
One control sets how much tonee changes in your set without checking with you first. It's a dial you turn, not a rule the model interprets. tonee never decides for itself that something is safe enough to skip the prompt.
01 /The four modes
Each mode is a line drawn through the same ladder the Undo Guard uses. Everything below the line runs silently; everything above it stops and asks. New chats start in Auto.
Read that as a table and it's easier to see where the lines fall. ⚠ stops and asks; · runs without a prompt.
02 /What a confirmation looks like
You get one prompt per change tonee makes, not one per parameter. A single card lists every write in that change, in plain language, with the real target names already resolved. Deletions are flagged, and anything that can't be undone says so on its own line.
The card never shows you code. What it shows is the list of things that are about to happen to your set, which is the part you can actually judge.
03 /Switching mode
Full Auto is deliberately harder to reach than the other three. It doesn't appear in the menu until you tick Enable Full Auto in Settings → Behavior, and even then it can't be set as your default. You choose it for one chat, and the next chat you open is back on your normal default.
04 /What never prompts
The gate only sits in front of writes. Anything that reads your set runs at full speed in every mode, Ask included. Otherwise the assistant couldn't answer a question without a dialog first.
The mode is read by the app, not by the model. Asking tonee to "stop confirming things" won't change it. That's the point. Use the ⋯ menu.