ElevenLabs Speech-to-Text

Transcribe audio files using ElevenLabs Speech-to-Text (Scribe v2).

Security Scan

The skill is internally consistent with its stated purpose (uploads audio to ElevenLabs Scribe v2 using ELEVENLABS_API_KEY); the only notable issue is a minor tooling omission (the script uses jq but the declared required binaries list only curl).

Purpose & Capability

The skill's name, description, required env var (ELEVENLABS_API_KEY), and the included script all match the stated purpose of calling ElevenLabs Speech-to-Text (Scribe v2). Minor mismatch: SKILL metadata and registry requirements list only curl as a required binary, but the script also invokes jq for JSON parsing (and README lists jq as 'optional'). Because the script uses jq in normal code paths, jq is effectively required.

Instruction Scope

Runtime instructions and the script stay within scope: they read a local audio file, require ELEVENLABS_API_KEY, and POST the file to the ElevenLabs API endpoint (https://api.elevenlabs.io/v1/speech-to-text). The script does not attempt to read unrelated files, other environment variables, or contact unexpected external endpoints.

Install Mechanism

No install spec is provided (instruction-only with a helper script). Nothing is downloaded or written by an installer; risk from install mechanism is low.

Credentials

Only ELEVENLABS_API_KEY is requested and declared as the primary credential, which is appropriate for a speech-to-text integration. No unrelated secrets or config paths are requested.

Persistence & Privilege

The skill does not request permanent presence (always:false) or modify other skills or system-wide settings. Default autonomous invocation is allowed (platform default) but not combined with other concerning permissions.

Assessment

This skill appears to do exactly what it claims: it sends a local audio file to ElevenLabs' Speech-to-Text API using the ELEVENLABS_API_KEY. Before installing or running it, consider the following: (1) Provide an API key with appropriate permissions and be aware that audio will be uploaded to ElevenLabs (privacy/billing implications). (2) The script uses jq for JSON handling even though only curl is declared as required — install jq or update the script so it doesn't fail (the script uses set -euo, so missing jq will cause it to exit). (3) Review the script yourself if you have sensitive audio; running it in an isolated environment is prudent. (4) If you want stricter control, verify the exact API endpoint and header usage against ElevenLabs' official docs and limit key scope/rotation as needed.

Transcribe audio files using ElevenLabs' Scribe v2 model. Supports 90+ languages with speaker diarization.

Quick Start

# Basic transcription
{baseDir}/scripts/transcribe.sh /path/to/audio.mp3

# With speaker diarization
{baseDir}/scripts/transcribe.sh /path/to/audio.mp3 --diarize

# Specify language (improves accuracy)
{baseDir}/scripts/transcribe.sh /path/to/audio.mp3 --lang en

# Full JSON output with timestamps
{baseDir}/scripts/transcribe.sh /path/to/audio.mp3 --json

Options

Flag	Description
--diarize	Identify different speakers
--lang CODE	ISO language code (e.g., en, pt, es)
--json	Output full JSON with word timestamps
--events	Tag audio events (laughter, music, etc.)

Supported Formats

All major audio/video formats: mp3, m4a, wav, ogg, webm, mp4, etc.

API Key

Set ELEVENLABS_API_KEY environment variable, or configure in clawdbot.json:

{
  skills: {
    entries: {
      "elevenlabs-stt": {
        apiKey: "sk_..."
      }
    }
  }
}

Examples

# Transcribe a WhatsApp voice note
{baseDir}/scripts/transcribe.sh ~/Downloads/voice_note.ogg

# Meeting recording with multiple speakers
{baseDir}/scripts/transcribe.sh meeting.mp3 --diarize --lang en

# Get JSON for processing
{baseDir}/scripts/transcribe.sh podcast.mp3 --json > transcript.json