byted-voice-to-text

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill explicitly accepts and fetches arbitrary public audio URLs (see asr_flash.py and asr_standard.py --url and the routing_strategy.md "公网 URL" rules) and instructs the agent to treat the ASR transcription as the user's message, so untrusted third-party audio can be ingested and materially drive agent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). Yes — the ensure_ffmpeg.py script (invoked at runtime when ffmpeg/ffprobe are missing) may run a repo-fallback command that downloads and installs an RPM from https://mirrors.rpmfusion.org/free/el/rpmfusion-free-release-{rhel_major}.noarch.rpm, which fetches and executes remote code to provision a required dependency (ffmpeg).

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill requires the agent to autonomously run an installation script (ensure_ffmpeg.py --execute) that may install system-level binaries (ffmpeg/ffprobe) and thus modify the machine state and potentially require elevated privileges.