Voice Transcriber Pro

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

This skill is functionally a transcriber, but the metadata, docs, and code don't line up. Before installing: (1) Do not provide DEEPGRAM_API_KEY expecting it to be used — the scripts currently call OpenAI/openrouter endpoints instead. (2) Inspect ~/.openclaw/workspace/.env for secrets; the transcribe.sh will read that file for OPENROUTER_API_KEY if present, which could expose other keys. (3) If you want Deepgram, either modify the scripts to actually use Deepgram or prefer a skill that declares and uses the correct provider. (4) Be aware that audio files will be uploaded to external services (api.openai.com and openrouter.ai) — only do this if you are comfortable sending sensitive audio to those providers. (5) If you lack the ability to audit/modify the code, run this only in a controlled/sandboxed environment and avoid placing other credentials in ~/.openclaw/workspace/.env. Given the mismatches and undeclared env usage, treat this skill with caution or seek a corrected/reviewed release.

Type: OpenClaw Skill Name: voice-transcriber-pro Version: 1.0.0 The skill is classified as suspicious primarily due to a path traversal vulnerability in `scripts/save_voice_note.py`. The `shutil.copy2(audio_path, audio_dest)` call does not sanitize the `audio_path` input, allowing an attacker to write files to arbitrary locations on the filesystem by providing paths like `../../evil.ogg`. Additionally, the `transcript` argument is written directly into a markdown journal file without sanitization, creating a potential prompt injection vector if the agent later processes its own journal entries. The `scripts/transcribe.sh` script performs expected API calls to OpenAI/OpenRouter for transcription and handles API keys in a standard manner for OpenClaw skills, without clear malicious intent.