Skill tab-complete unexpectedly steals focus while navigating large input messages

What version of Codex CLI is running?

codex-cli 0.136.0

What subscription do you have?

Pro

Which model were you using?

gpt-5.5

What platform is your computer?

Darwin 25.5.0 arm64 arm

What terminal emulator and version are you using (if applicable)?

iterm2

Codex doctor report

What issue are you seeing?

When editing a multi-line prompt that contains custom skill references, moving the cursor with the arrow keys across an already accepted skill token causes the autocomplete TUI prompt to reappear.

For example, draft a long multi-paragraph message and include skill tokens such as $create-pr or $review-loop (whatever ones you actually have accessible from your own custom skills though). After accepting the autocomplete suggestion, the token is decorated cyan as expected. However, if you later use the arrow keys to move the cursor back across that token while editing the message, Codex shows the autocomplete prompt again and requires Enter or Tab to dismiss/accept it.

This happens even though the skill token was already accepted and visually marked as accepted.

Expected behavior:

Once a skill autocomplete has been accepted, moving the cursor across that completed/decorated token should not reopen the autocomplete prompt.

Actual behavior:

The autocomplete prompt reopens when the cursor crosses the accepted skill token, interrupting normal cursor navigation and requiring an extra Enter or Tab to continue editing.

What steps can reproduce the bug?

See main report

What is the expected behavior?

See main report.

Additional information

_No response_