[BUG] Japanese characters displayed as octal escape sequences in @ file mention preview

Environment

• Claude Code version: 2.1.6
• OS: Windows 11 + WSL2 (Linux 6.6.87.2-microsoft-standard-WSL2)
• Terminal: Windows Terminal
• Locale: ja_JP.UTF-8

Description

When using the @ file mention feature in Claude Code CLI, Japanese (and likely other CJK/multibyte UTF-8) characters in file paths are displayed as octal escape sequences instead of the actual characters.

Steps to Reproduce

1. Create a file with Japanese characters in the filename, e.g.:

``
touch "Gmailスパム判定の仕組み.md"
``

1. In Claude Code CLI, type @ and start typing the file path
2. Observe the file preview/suggestion

Expected Behavior

The file path should display Japanese characters correctly:

@10_Notes/20260108_1230_Gmailスパム判定の仕組み_送信者レピュテーションが決め手.md

Actual Behavior

Japanese characters are displayed as octal escape sequences:

@""10_Notes/20260108_1230_Gmail\343\
  202\271\343\203\221\343\203\240\345\
  210\244\345\256\232\343\201\256\344\
  273\225\347\265\204\343\201\277_\351
  \200\201\344\277\241\350\200\205\343
  \203\254\343\203\224\343\203\245\343
  \203\206\343\203\274\343\202\267\343
  \203\247\343\203\263\343\201\214\346
  \261\272\343\202\201\346\211\213.md"

Technical Analysis

The octal sequences like \343\202\271 correspond to UTF-8 bytes:

• \343 = 0xE3 (first byte of 3-byte UTF-8 character)
• \202 = 0x82
• \271 = 0xB9
• Together: E3 82 B9 = "ス" (Katakana SU)

This suggests the Rust code is using Debug formatting ({:?}) or similar byte-level representation instead of proper UTF-8 string display for the file path preview.

Possible Fix Location

The issue is likely in the @ file autocomplete/preview rendering code where file paths are formatted for terminal display. The code should use Display formatting ({}) or proper UTF-8 string handling instead of Debug formatting.