User's own pasted input is folded as '(N lines hidden)' and cannot be expanded (Ctrl-O) or copied back out of the transcript

Summary

When the user pastes a long block of text into the Claude Code prompt, the model receives it in full, but the transcript view folds the middle of it behind a placeholder such as ──── (108 lines hidden) ────. The user cannot get that text back from the UI:

• Ctrl-O does not reveal the hidden lines of the user's own paste;
• copying from the terminal copies the placeholder, not the text;
• so the user can no longer see or retrieve content they typed, which the model can see and is acting on.

This is a terminal tool used for exact work on code and documents. The user's own input should always be fully viewable and copyable in the transcript. Nothing about a terminal requires folding it.

Environment

• Claude Code 2.1.288, Linux terminal (Debian), interactive session
• Paste of about 150–170 lines of Markdown from Sublime Text, including long lines and Arabic text

Steps to reproduce

1. Paste about 150+ lines of text into the prompt and send it.
2. Scroll back to the user message in the transcript: its middle is folded as ──── (N lines hidden) ────.
3. Press Ctrl-O (expand): the hidden lines of the user message are still not shown.
4. Select and copy the message from the terminal: the clipboard contains the placeholder and the display padding (indentation, padding to terminal width), not the original text.

Side effect

If the user then re-pastes that copied text, the model receives the placeholder line in place of the content. The model works from incomplete input, and the cause is invisible to both.

Expected

• The user's own messages are always viewable in full in the transcript, at least through Ctrl-O or another expand action.
• Copying a user message from the transcript yields the original text, without folding, placeholders or display padding.
• If folding is kept for brevity, it is a display toggle on the user's own input that can always be undone.