Voice2text

Offline speech-to-text conversion using Vosk local model; input audio file path, output transcript text.

This skill appears internally consistent, but it runs publisher-provided Python code locally — review and run it in an isolated environment. Before using: 1) ensure you have a compatible 16kHz mono 16‑bit WAV and a trusted Vosk model placed in a 'model' directory (the package does not include the model); 2) install dependencies from requirements.txt in a virtualenv; 3) inspect main.py yourself (it is short and straightforward) and avoid running untrusted models; 4) if you need stronger isolation, run inside a container or VM. The tests are placeholders and do not create sample audio, so functional testing requires providing a real WAV and a Vosk model.

Type: OpenClaw Skill Name: voice2text Version: 1.0.0 The skill bundle implements an offline speech-to-text function using the Vosk library. Its core functionality involves opening a user-provided audio file path (`audio_path`) using `wave.open()` in `main.py`. While user-controlled file paths can be a vulnerability, the `wave.open()` function is specific to WAV files and is highly unlikely to successfully open and parse arbitrary non-WAV files (e.g., `/etc/passwd`) to leak their content; it would typically raise an error. There is no evidence of intentional malicious behavior such as data exfiltration, command execution, persistence, or prompt injection attempts against the AI agent. The file access is inherent to the skill's stated purpose, and the specific library function used mitigates the risk of arbitrary file reads.