Qwen Audio
Security checks across malware telemetry and agentic risk
Overview
The skill’s audio features match its stated purpose, but it can automatically install a prerelease Python package during use and it stores reusable voice samples locally.
Review this skill before installing if you do not want runtime package installation. Use an isolated Python environment, prefer pinned dependencies, and only provide voice samples you have permission to use and are comfortable storing locally.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
On macOS, a TTS request could download and install package code automatically if the dependency is missing.
A normal runtime path can invoke a shell command that modifies the Python environment and installs a prerelease dependency, rather than limiting installation to an explicit setup step.
except ImportError:
print("mlx-audio 未安装，正在安装...", file=sys.stderr)
os.system("uv add mlx-audio --prerelease=allow")Move dependency installation to an explicit install step, ask the user before installing packages at runtime, and pin or lock dependency versions.
Different installs may resolve to different package versions, and users rely on the upstream package sources being trustworthy.
These external ML dependencies are expected for the skill, but several are unpinned or version-ranged, which increases supply-chain variability.
dependencies = [
"mlx-audio>=0.3.1; platform_system == 'Darwin'",
"qwen-asr; platform_system != 'Darwin'",
"qwen-tts>=0.1.1; platform_system != 'Darwin'",
"torch; platform_system != 'Darwin'"
]Use a lockfile or pinned versions for reproducible installs, and install in an isolated environment.
Voice samples and transcripts may remain on disk after use and could be reused in later tasks.
The skill persists reusable voice profiles, including reference audio and transcripts, for later use.
Voices are stored in the `./voices/` directory at the skill root level. Each voice has its own folder containing: - `ref_audio.wav` - `ref_text.txt` - `ref_instruct.txt`
Only store voice samples you are comfortable keeping locally, delete unused voice profiles, and avoid using sensitive recordings unless needed.
Generated audio could be mistaken for a real person if used without consent or disclosure.
Voice cloning is clearly disclosed and purpose-aligned, but it can create convincing synthetic speech that may be misused for impersonation.
Clone any voice using a reference audio sample. Provide the wav file and its transcript
Use voice cloning only with permission, keep source recordings private, and label generated audio as synthetic when sharing it.