Japanese (UTF-8) text file attachments are garbled in chat, but read correctly via the Read tool (path-based)

Status Open
Maintainer reply None cached
Activity 0 comments · opened Aug 18, 2026

Environment: Claude Code, VS Code extension, macOS

Description:
When a UTF-8-encoded text file containing Japanese text is uploaded as a chat attachment, its content appears garbled — consistent with UTF-8 bytes being misinterpreted as Latin-1/Windows-1252. The exact same file, when read via the Read tool using its full file path, displays correctly.

Steps to reproduce:

  1. Create a UTF-8 text file with Japanese content, e.g.:

``
これはテストです。
文字化けの再現確認用サンプルです😀
タイムスタンプ:2026-08-19 12:00
``

  1. Attach/upload this file directly in the chat.
  2. Observe the rendered content — it is garbled:

``
ããã¯ãã¹ãã§ãã
æå­åãã®åç¾ç¢ºèªç¨ãµã³ãã«ã§ãð
ã¿ã¤ã ã¹ã¿ã³ãï¼2026-08-19 12:00
``

  1. Instead, ask Claude to read the same file via its full path (e.g., /Users/xxx/Downloads/test.txt).
  2. Observe that the content now displays correctly (matching the original).

Expected: Chat-attached UTF-8 text files should display the same correct content as path-based Read access.

Actual: Chat-attached file content is garbled; path-based Read access is correct.

Additional notes: Also reproduced with a real-world UTF-8 text file (Discord chat log export, ~1500 lines with Japanese text and emoji) — same garbling pattern when attached, correct when read via path.

View original on GitHub ↗