[BUG]

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?

I have built an agent that should create slackposts from my audio when I hit CAPSLOCK and speak to Claude. I have pasted the conversation I had from terminal which should explain the issue:

/slackpost

⏺ I need your voice memo or spoken thoughts to format into a Slack post. What would you like to share?

❯ I have typed /slackpost and pressed capslock which activated claude's built in audio recording. After I hit capslock again and pressed sent, it opened the claude
desktop and no message formatting took place as I would've expected

⏺ That sounds like unexpected behavior — the audio transcription should have fed into the /slackpost prompt as text input, not opened a separate window.

A couple of things that may have happened:

• The capslock audio feature might have routed the recording to a new Claude Desktop session rather than inserting the transcription back into the current CLI prompt
• The /slackpost skill loaded and was waiting for input, but the audio result landed somewhere else

To work around it now: Just paste or type your message here and I'll format it for Slack.

For the bug: You can report it at anthropics/claude-code — include the steps you described (typing a skill command, then using capslock audio recording) so the team
can reproduce it.

❯ Can you please check the code and see if you can fix it?

Searched for 2 patterns, read 1 file (ctrl+o to expand)

⏺ The skill file looks fine — the issue isn't in the skill itself. Here's what's happening:

The skill uses $ARGUMENTS to receive input, which means it expects the content to come inline with the command, like:

/slackpost <your transcribed text here>

When you type /slackpost and then press Capslock to record after, the command gets submitted first with empty $ARGUMENTS, and the audio transcription ends up
disconnected from it.

The fix needs to be in the Claude Code app (how Capslock audio interacts with slash commands that are already typed) — that's outside what I can change from the skill
file.

Workaround in the meantime:

1. Press Capslock first to record your audio
2. Once the transcription appears in the input box, prepend /slackpost before it
3. Then send

What Should Happen?

the audio transcription should have fed into the /slackpost prompt as text input, not opened a separate window.

Error Messages/Logs

Steps to Reproduce

Build any command that should intake audio from speaking to claude.

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

Claude Code 2.1.91

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_