Voice Transcriber

Voice note transcription and archival for OpenClaw agents. Powered by Deepgram Nova-3. Transcribes audio messages, saves both audio files and text transcript...

Security Scan

The skill claims Deepgram-based transcription but its scripts call OpenAI/OpenRouter endpoints and read other agent config files while requiring a DEEPGRAM_API_KEY that is never used — the pieces are internally inconsistent and warrant caution.

Purpose & Capability

The skill description and SKILL.md advertise Deepgram Nova-3, and the declared primary credential is DEEPGRAM_API_KEY, but the shipped scripts call OpenAI (Whisper) and OpenRouter endpoints instead. This mismatch indicates the declared purpose/credentials do not align with the implementation.

Instruction Scope

Runtime instructions direct the agent to run two scripts. The transcribe.sh script reads an additional environment variable (OPENROUTER_API_KEY) and attempts to use OPENAI_API_KEY, neither of which are declared in requires.env. It also attempts to read ~/.openclaw/workspace/.env for keys. save_voice_note.py writes audio and transcripts into ~/.openclaw/workspace/memory (expected for a journaling skill), but the transcribe script accesses configuration outside the declared scope.

Install Mechanism

No install spec is provided (instruction-only with small scripts), so there is no remote code download or archive extraction risk. The scripts will run locally when invoked.

Credentials

The registry metadata requires DEEPGRAM_API_KEY as the primary credential, but neither script references DEEPGRAM_API_KEY. Instead they rely on OPENAI_API_KEY and OPENROUTER_API_KEY (not declared). This is disproportionate and inconsistent: either the metadata is wrong or the code was repurposed without updating env requirements. The skill also references reading a local ~/.openclaw/workspace/.env file to discover keys, which effectively accesses other stored credentials.

Persistence & Privilege

The skill does not request 'always' or other elevated platform privileges. It writes files to ~/.openclaw/workspace/memory (audio and journal entries) which is expected for a voice-journaling skill, but reading ~/.openclaw/workspace/.env to obtain keys is a cross-cutting access to agent configuration and should be considered sensitive.

What to consider before installing

Do not assume this skill uses Deepgram despite the description and required DEEPGRAM_API_KEY — the scripts call OpenAI/OpenRouter and read other agent config files. Before installing or providing credentials: (1) inspect and, if needed, fix the scripts so the credential names match the service you intend to use; (2) avoid handing over DEEPGRAM_API_KEY, OPENAI_API_KEY, or OPENROUTER_API_KEY unless you trust the author and have verified where they will be sent; (3) note the skill writes audio and transcripts to ~/.openclaw/workspace/memory — sensitive audio will be stored there; (4) consider running the scripts in a sandbox or isolated account, or request a corrected package from the owner that declares the actual required env vars and removes reading ~/.openclaw/workspace/.env. If you want, ask the author to clarify whether Deepgram is supported and to update metadata to match the implementation.

Audio transcription via Whisper + voice note journaling.

When to Use

• Transcribing voice messages or audio files
• Saving voice notes with full transcripts
• Converting speech to text from any audio format

Usage

Transcribe audio

bash {baseDir}/scripts/transcribe.sh /path/to/audio.ogg
bash {baseDir}/scripts/transcribe.sh /path/to/audio.ogg --out /path/to/output.txt

Save voice note with transcript

python3 {baseDir}/scripts/save_voice_note.py /path/to/audio.ogg "Optional context"

Supported Formats

OGG, MP3, WAV, M4A, FLAC, WEBM

Credits

Built by M. Abidi | agxntsix.ai YouTube | GitHub Part of the AgxntSix Skill Suite for OpenClaw agents.

📅 Need help setting up OpenClaw for your business? Book a free consultation