audio

Audio Production

Generate audio — from single-line TTS to multi-voice podcasts. Currently powered by ElevenLabs. Includes ready-to-run Python scripts. Claude writes them to a temp file and executes directly.

Boundaries

This skill MAY: install the SDK, generate audio files, list voices, create voice clones, write and execute generation scripts, play audio.

This skill MAY NOT: store API keys in code (use env vars or ~/.elevenlabs/api_key), commit audio files to git, generate audio without user approval of the script first.

Common Rationalizations

Phase 0: Environment Setup

Entry: User wants audio content.

Step 1: Detect Authentication

Check in order — use the first one found:

# 1. Check CLI auth (preferred — already logged in)
elevenlabs auth whoami --no-ui 2>/dev/null

# 2. Check env var
echo "${ELEVENLABS_API_KEY:+set}"

# 3. Check stored key file
cat ~/.elevenlabs/api_key 2>/dev/null | head -c 10

API key resolution in Python (use this in ALL scripts):

import os

def get_api_key() -> str:
    """Resolve ElevenLabs API key from CLI store, env var, or fail."""
    key_file = os.path.expanduser("~/.elevenlabs/api_key")
    if os.path.exists(key_file):
        return open(key_file).read().strip()
    key = os.environ.get("ELEVENLABS_API_KEY", "")
    if key:
        return key
    raise RuntimeError(
        "No ElevenLabs API key found. Run `elevenlabs auth login` or "
        "set ELEVENLABS_API_KEY environment variable."
    )