voice-promptlisted
Install: claude install-skill jayu1023/voice-prompt-skill
# voice-prompt — Speak to Claude Code
Turns your microphone into a hands-free input for Claude Code on macOS.
- **Records** continuously from the default mic
- **Transcribes** each chunk locally with faster-whisper (no API calls, no torch)
- **Types** the words into whichever window is focused (your Claude Code terminal)
- **Submits** the prompt when you say a trigger phrase: `go`, `send it`, `hit enter`, `submit`, `enter`
- **Stops** when you say: `stop listening`, `quit`, or `exit`
## When this skill is invoked
If the user is asking to **set up / install** voice input, run the setup flow:
1. Check whether deps are installed:
```bash
python3 -c "import sounddevice, faster_whisper, numpy" 2>/dev/null && echo OK || echo MISSING
```
2. If `MISSING`, run the setup script and stream output back to the user:
```bash
bash ~/.claude/skills/voice-prompt/setup.sh
```
3. Remind the user to grant **Microphone** and **Accessibility** permissions to their terminal app (Terminal.app, iTerm2, etc.) in **System Settings → Privacy & Security**. Without Accessibility, the typing-via-keystroke step will silently no-op.
If the user is asking to **start listening / dictate now**, launch the listener in a new Terminal window so it can speak into the *current* Claude Code session:
```bash
bash ~/.claude/skills/voice-prompt/launch.sh
```
This opens a separate Terminal window running the listener. The user clicks back into the Claude Code window, speaks, and watches their wo