`/permission` selector: descriptions for "Auto" and "YOLO" are swapped

Summary

The one-line descriptions shown for Auto and YOLO in the /permission selector do not match how the official documentation defines these two modes. As shown, Auto reads as the more dangerous mode, while the docs describe YOLO as the one that skips approvals for everything. This is misleading when choosing a permission mode.

What the TUI shows

Manual                                                                                                                                                                                      
  Approve every action yourself.
Auto                                           
  Run all actions automatically, including risky ones.
YOLO                                           
  AI decides which actions need your approval.

These strings come from PERMISSION_OPTIONS in the permission selector component, and the Auto string is also reused in the /auto toggle notice (Auto mode: ON — Run all actions automat ically, including risky ones.).

What the official docs say

• YOLO is the mode that bypasses approvals:
  • "YOLO mode (/yolo) skips the approval confirmation for almost all tool calls" — Interaction and input, whic
    h also warns: "YOLO mode skips confirmation for file writes and command execution. Only use it in working directories you trust."
  • default_permission_mode: "yolo (auto-approve everything)" — Configuration files
• Auto is the restrained one:
  • "Auto mode (/auto) is more restrained: tool approvals are handled automatically, but the agent does not ask the user clarifying questions — useful when you want unattended operation
    without fully disabling approvals." — Interaction and input
  • default_permission_mode: "auto (auto-approve read operations)" — Configuration files
  • /auto: "tool approvals are handled automatically and the Agent will not ask the user questions" — [Slash commands](https://www.kimi.com/code/docs/en/kimi-code-cli/reference/slash-com
    mands.html)

The problem

• "Run all actions automatically, including risky ones." describes the documented behavior of YOLO, not Auto.
• "AI decides which actions need your approval." does not match YOLO at all — in YOLO almost nothing needs your approval. It reads closer to the documented Auto behavior.
• Net effect: the menu implies Auto is riskier than YOLO, inverting the documented risk ordering (manual < auto < yolo). A user choosing a mode based on these descriptions may enable one t
  hey did not intend.

Suggestion

Please review these two description strings (and the /auto toggle notice that reuses the Auto string) and align them with the documented semantics — for example:
(Exact wording up to the team; the key point is that the copy should reflect that YOLO carries the highest risk.)

Environment

• kimi --version: 0.26.0 (Linux x86_64)

• Also confirmed present in the latest release at the time of writing, 0.27.0 (npm package @moonshot-ai/kimi-code@0.27.0, dist/main.mjs); the 0.27.0 changelog lists no related fix.

• Auto: "Approve routine actions automatically; the agent won't pause to ask you questions."

• YOLO: "Skip all approvals and run everything, including risky actions."

Generated by Kimi K3

Start at the permission selector component and inspect PERMISSION_OPTIONS, then check the /auto toggle notice that reuses the Auto description. Compare both strings with the documented Auto and YOLO semantics, and confirm the selector and toggle show restrained Auto behavior and approval-skipping YOLO behavior.