## 🚨 CRITICAL: Rust panic when reading files with Japanese characters in textarea

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?

### Summary
Claude Code crashes with a fatal runtime error when reading PHP files containing Japanese text within textarea elements. This causes complete session termination and data loss.

### Error Message
thread '' panicked at /rustc/.../library/core/src/str/mod.rs:833:21:
byte index 1 is not a char boundary; it is inside 'け' (bytes 0..3) of け'); ?></textarea>
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
fatal runtime error: failed to initiate panic, error 5, aborting
Aborted (core dumped)

### Environment

• OS: WSL2 (Linux 6.6.87.2-microsoft-standard-WSL2)
• Claude Code version: Latest
• File encoding: UTF-8

### Steps to Reproduce

1. Create a PHP file with Japanese text inside a textarea default value:

```php
<textarea><?php echo h($var ?? '日本語テキスト逝け！'); ?></textarea>

1. Run Claude Code in the directory
2. Attempt any operation that reads the file

Impact

• Severity: Critical
• Frequency: 100% reproducible
• Data Loss: Session terminates abruptly, unsaved work lost
• Workaround: None - cannot work with files containing this pattern

Root Cause

Rust string slicing at byte index 1 fails because Japanese character 'け' occupies bytes 0-3. The code incorrectly assumes ASCII byte boundaries.

Expected Behavior

Claude Code should handle UTF-8 multi-byte characters correctly without crashing.

Suggested Fix

Use char_indices() or grapheme_indices() instead of direct byte indexing when processing string content.

---日本語が問題みたいですが、すぐにオチて使い物にならなくなってきた至急対策を！

What Should Happen?

途中で停止するのは辞めてください！すぐ落ちて何度も何度も立ち上げなおさないといけないです。途中でクラッシュしないように！

Error Messages/Logs

thread '<unnamed>' (617546) panicked at /rustc/ed61e7d7e242494fb7057f2657300d9e77bb4fcb/library/core/src/str/mod.rs:833:21:
  byte index 1 is not a char boundary; it is inside 'け' (bytes 0..3) of `け'); ?></textarea>`
  note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
  fatal runtime error: failed to initiate panic, error 5, aborting
  Aborted (core dumped)



 Trigger file: admin/settings.php
  Trigger pattern: Japanese text inside PHP textarea default value
  Example: <textarea><?php echo h($var ?? '...逝け！'); ?></textarea>
  The character 'け' (U+3051, 3 bytes in UTF-8) causes byte boundary violation when sliced at index 1.

Steps to Reproduce

10分程度でおちます。

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

2.0.76

Claude Code Version

2.0.76

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

### Environment Details

• OS: WSL2 on Windows (Linux 6.6.87.2-microsoft-standard-WSL2)
• Shell: Bash
• File System: NTFS via /mnt/c/ (Windows path mounted in WSL)
• File Encoding: UTF-8 (verified with file -i)

### Reproduction Steps

1. Create a PHP file with this content:

```php
<textarea><?php echo h($value ?? '今すぐ逝け！'); ?></textarea>

1. Navigate to the directory containing this file
2. Run any Claude Code command that triggers file reading (e.g., search, edit, or exploration)
3. Claude Code crashes immediately

Minimal Reproducible Example

<?php
// test.php - Minimal crash trigger
?>
<textarea><?php echo '逝け'; ?></textarea>

Observations

• Crash occurs during file content parsing, not during explicit file read
• The specific character 'け' (hiragana KE, U+3051) is mentioned in error
• Panic occurs in Rust's core string module (str/mod.rs:833)
• Error 5 in panic handler suggests memory/thread issue

Workaround Attempted

• None successful - avoiding the file entirely is the only option
• Cannot use --exclude patterns as crash happens before command processing

Frequency

• 100% reproducible
• Affects all operations in directories containing such files