[BUG] Korean text with certain trailing characters gets corrupted to "[Image #N]" format when clipboard contains image

Environment

• Platform (select one):
• [x] Claude Code CLI
• [ ] Anthropic API
• [ ] AWS Bedrock
• [ ] Google Vertex AI
• [ ] Other: <!-- specify -->
• Claude CLI version: 1.0.72
• Operating System: <!-- e.g. macOS 14.3, Windows 11, Ubuntu 22.04 -->
• Terminal: <!-- e.g. iTerm2, Terminal App -->

Bug Description

Korean text followed by specific characters (comma, numbers, period) gets unexpectedly converted to "[Image #N]" format during text input or processing in Claude Code CLI. This issue only occurs when there is an image present in the clipboard.

https://github.com/user-attachments/assets/328da3b9-cccf-49e6-b6e6-60b2de5a98c1

Steps to Reproduce

1. Copy an image to clipboard (e.g., screenshot or image file)
2. Type Korean text followed by any of these characters:

• Comma: 안녕, → becomes 안[Image #N]
• Numbers: 안녕0 → becomes 안[Image #N]
• Period: 안녕. → becomes 안[Image #N]
• Numbers: 안녕1 → becomes 안[Image #N]

Expected Behavior

Korean text with trailing punctuation or numbers should remain unchanged regardless of clipboard contents:

• Input: 안녕, → Expected: 안녕,
• Input: 안녕0 → Expected: 안녕0
• Input: 안녕. → Expected: 안녕.
• Input: 안녕1 → Expected: 안녕1

Actual Behavior

When clipboard contains an image, Korean text gets partially replaced with image placeholder syntax:

• 안녕, → 안[Image #N]
• 안녕0 → 안[Image #N]
• 안녕. → 안[Image #N]
• 안녕1 → 안[Image #N]

Additional Context

• This issue only manifests when there is an image in the clipboard
• The corruption consistently replaces the last Korean character + trailing character with image reference syntax
• Suggests the CLI is incorrectly detecting clipboard image content and applying image insertion logic to Korean text + punctuation combinations
• May be related to how the CLI handles multi-byte Korean characters in combination with ASCII punctuation/numbers when clipboard monitoring is active