fix(web): clarify auto permission fallback

[t3-code]

what changed

Clarify that Auto only approves routine actions when the selected provider supports that review mode. Other providers continue to ask.

why

The previous copy implied every provider has an AI reviewer. Cursor and Grok currently fall back to supervised approvals in Auto mode.

ui changes

Text-only change in the permission mode menu. Visual evidence is pending in this draft because the runner exhausted its process limit while launching the real app.

checks

• web typecheck
• targeted formatting
• git diff --check

Model: OpenAI (API model not exposed). Harness: T3bot.

Note

Clarify 'Auto' runtime mode description in chat composer

Updates the UI description for the 'Auto' runtime mode in ChatComposer.tsx to say 'Supported providers approve routine actions; others still ask.' instead of 'An AI reviewer approves routine actions; risky ones still ask.' This better reflects that fallback behavior is determined by provider support, not an AI reviewer.

^{Macroscope summarized 6c0e4bf.}

---

Note

Low Risk
Copy-only change in the permission mode menu; no runtime or approval logic is modified.

Overview
Updates the Auto runtime mode helper text in the chat composer permission menu so it matches how approvals actually work.

The description now says supported providers approve routine actions and others still ask, instead of referring to a generic AI reviewer and risky actions. This aligns the UI with provider-specific behavior (e.g. Cursor and Grok staying supervised in Auto) without changing any approval logic.

^{Reviewed by Cursor Bugbot for commit 6c0e4bf. Bugbot is set up for automated code reviews on this repo. Configure here.}