The current approval policy "never" basically rejects any request that is not allowed by the sandbox. It should default to on-request just like what OpenAI's auto setting does so we can approve manually if needed.
OpenAI doc:
| Key |
Type / Values |
Details |
| --full-auto |
boolean |
Shortcut for low-friction local work: sets --ask-for-approval on-request and --sandbox workspace-write. |
Relevant code
Android:
|
AUTO( |
|
title = "Auto", |
|
description = "No prompts and workspace-write sandbox", |
|
approvalPolicy = "never", |
|
sandboxMode = "workspace-write", |
|
), |
iOS:
|
var description: String { |
|
switch self { |
|
case .readOnly: return "Ask before commands and run in read-only sandbox" |
|
case .auto: return "No prompts and workspace-write sandbox" |
|
case .fullAccess: return "No prompts and danger-full-access sandbox" |
|
} |
|
} |
|
|
|
var approvalPolicy: String { |
|
switch self { |
|
case .readOnly: return "on-request" |
|
case .auto, .fullAccess: return "never" |
|
} |
|
} |
The current approval policy "never" basically rejects any request that is not allowed by the sandbox. It should default to on-request just like what OpenAI's auto setting does so we can approve manually if needed.
OpenAI doc:
Relevant code
Android:
litter/apps/android/app/src/main/java/com/litter/android/ui/LitterAppShell.kt
Lines 5549 to 5554 in df016c7
iOS:
litter/apps/ios/Sources/Litter/Views/ConversationView.swift
Lines 1851 to 1864 in df016c7