Tested on WSL2 + Windows Terminal. Ctrl+Enter works correctly for inserting newlines after this patch. Note for Windows Terminal users: Shift+Enter sends CR (same as Enter) rather than LF, so Shift+Enter still submits -- Ctrl+Enter is the reliable shortcut on that platform until Kitty keyboard protocol is fully negotiated.

Would be worth updating the PR description or a follow-up comment to note the platform difference in case other WSL users find this PR.

I ended up creating a skill to auto-reapply this patch after every Hermes update, been three weeks and this is still open.
It works perfectly on Warp/macOS. Would love to see it merged so I can drop the local workaround.

I built a small follow-up branch on top of this PR that keeps the CR/LF preservation from #18229 but gates the raw-LF newline fallback so it doesn't turn every bare LF into multiline intent in thin/local PTY environments where plain Enter may also arrive as LF.

Branch / compare:
morozsm:fix/tui-lf-shift-enter-newline...dso2ng:fix/tui-modified-enter-intent

What it adds:

• keeps explicit modified Return (Shift+Enter, Ctrl+Enter, macOS action-mod Return) as newline intent
• treats raw LF (\n) as newline intent only for environments where plain Enter is expected to remain distinct from LF, e.g. SSH/WSL/Windows Terminal and known terminal programs like WezTerm/Warp/Ghostty/Kitty
• keeps bare LF as submit intent for generic local POSIX/thin PTY fallback cases, avoiding the "Enter becomes dead / never submits" risk mentioned in the classic CLI handling
• adds textInputModifiedEnter.test.ts coverage for explicit modified Return, raw LF allowlisted terminals, SSH/WSL/WT, and the generic thin-PTY fallback

Local verification on the branch:

• npm --prefix ui-tui test -- --run packages/hermes-ink/src/ink/parse-keypress.test.ts src/__tests__/textInputModifiedEnter.test.ts -> 23 passed
• npm --prefix ui-tui run type-check -> passed
• npm --prefix ui-tui run build -> passed
• non-ASCII added-lines guard -> non_ascii_added_lines=0

Happy to turn this into a patch against this PR if that shape is useful; I avoided opening a competing PR since this PR already owns the root cause.