[BUG] Read tool drops Unicode Private Use Area characters (e.g., Powerline glyphs)

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

The Read tool fails to preserve Unicode Private Use Area (PUA) characters when reading and displaying file contents. Standard emoji render correctly, but PUA glyphs like the Powerline branch symbol (U+E0A0, byte sequence ee 82 a0) are silently dropped.

PUA characters are silently dropped. Standard emoji (e.g., 🍺, 💻) in the same file render correctly. A hex dump (xxd) confirms the bytes are present in the file, but they do not survive the Read → model → output pipeline.

What Should Happen?

All valid UTF-8 characters in a file, including Private Use Area codepoints, should be faithfully reproduced in the Read tool output.

Error Messages/Logs

Steps to Reproduce

1. Create a file containing a PUA character, e.g., the Powerline branch symbol U+E0A0:

``shell
$ echo -e "branch: \xee\x82\xa0 main" > /tmp/test-pua.txt
``

1. Use Claude Code's Read tool to read the file.
2. The PUA character is missing from the output — it appears as a blank space.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.58 (Claude Code)

Platform

Claude Code

Operating System

macOS

Terminal/Shell

Other

Additional Information

Terminal: Alacrity 0.17