[BUG] Permission prompt options require type + Enter instead of single keypress (regression)

Resolved 💬 4 comments Opened Feb 25, 2026 by michaelgrafwebdev Closed Mar 1, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Permission prompt option selection has changed behavior. Previously, pressing a number key (1, 2, or 3) on the permission prompt would immediately select and execute that option. Now, the user must type the number AND press Enter for the selection to register.

This appears to be a regression in the interactive prompt/TUI handling. The extra keystroke adds friction to every single permission interaction throughout a session.

Example Prompt

 Bash command
   cd /Users/user/Projects/my-project && npx jest --no-coverage 2>&1 | tail -8
   Full test suite still passes
 Do you want to proceed?
 ❯ 1. Yes
   2. Yes, and don't ask again for: tail:*
   3. No

Previous behavior: Pressing 2 immediately selects option 2 and proceeds.
Current behavior: Pressing 2 types the character but does not select. User must then press Enter to confirm.

This affects all permission prompts (bash commands, file edits, etc.), not just this specific example.

What Should Happen?

Pressing a number key (1, 2, 3) on the permission prompt should immediately select and execute the corresponding option, as it did in previous versions. No additional Enter keypress should be required.

Error Messages/Logs

No error messages — the prompt simply does not respond to the single keypress.

Steps to Reproduce

  1. Start a Claude Code session
  2. Trigger any action that requires a permission prompt (e.g., a bash command)
  3. When the numbered option list appears, press 2 (or any number)
  4. Observe that the option is not selected — the number is typed but not acted upon
  5. Press Enter to confirm — only then does the selection execute

Claude Model

Sonnet

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Not certain of exact version — behavior changed recently (noticed in v2.1.58)

Claude Code Version

2.1.58

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

  • Machine: Apple M1 MacBook Pro
  • The wildcard option (e.g., tail:*) on option 2 may be related to the behavior change, but the regression affects all numbered options including option 1 (plain "Yes") and option 3 ("No")

View original on GitHub ↗

This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗