[Bug] The "paste summary" causes the content to be disordered after pasting.

[zclllyybb]

Description

When UTF8 characters are included in the input content, if the paste action triggers a paste summary, the actual text passed to the model will be completely scrambled, and the pasted text will be inserted in the wrong position.

Plugins

none

OpenCode version

dev and any release version

Steps to reproduce

1. Enter some UTF8 text
2. Paste a long text to trigger paste summary, confirm that you see "[Pasted ~x lines]" displayed.
3. Send it. then you can see the disordered text

Screenshot and/or share link

what I actually entered is:

我能吞下玻璃而不伤身体我能吞下玻璃而不伤身体我能吞下玻璃而不伤身体[Pasted ~1 lines]我能吞下玻璃而不伤身体。

which the pasted line is:

A sufficiently long test content1 A sufficiently long test content2 A sufficiently long test content3 A sufficiently long test content1 A sufficiently long test content2 A sufficiently long test content4 A sufficiently long test content1 A sufficiently long test content2 A sufficiently long test content5 

Operating System

OS-independent

Terminal

terminal-independent

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• [Pasted ~N lines] is partially submitted when prompt contains CJK or other multi-width characters #17032: Describes the exact same root cause — paste summary replacement logic mixes terminal visual-width offsets with JavaScript string indices, causing garbled/misaligned content when CJK or other multi-width characters are present before a [Pasted ~N lines] token
• Text after paste content will be cut #21475: Text after paste content is cut when CJK characters precede the paste block
• TUI输入框在两个粘贴文本块之间输入文字，发送后会被吞掉 #22786: Text typed between two paste blocks is lost on submission (also CJK-related)

If your issue is the same underlying bug, consider adding a comment there instead.

[zclllyybb]

I have fixed this problem in #25855 and the tests runs well locally. @jlongster would you mind telling me how to continue to push this?