PureVocals-UVR-Automator 安全扫描报告 — 可信 | ClawSafe

0 /100

PureVocals-UVR-Automator

批量从音频文件中提取超干净干声（Vocals Only），支持 VR Architecture、自动 GPU 检测和虚拟环境管理

PureVocals-UVR-Automator is a legitimate audio vocal extraction skill. All subprocess/shell usage is directly tied to documented audio processing features (venv management, audio-separator CLI, ffmpeg downloader, pip installs). No credential theft, data exfiltration, obfuscation, or undeclared sensitive behavior was found.

技能名称PureVocals-UVR-Automator

分析耗时37.6s

引擎pi

✓

可以安装

Skill is safe to use. Consider pinning dependency versions in a requirements.txt for reproducible builds.

安全发现 2 项

严重性	安全发现	位置
低危	Dependency versions not pinned 供应链 Dependencies such as pydub, audio-separator, librosa, huggingface-hub are installed without version constraints. While common in development scripts, this could lead to unexpected behavior if a dependency publishes a breaking release. `ensure_package.pip("audio-separator")` → Add a requirements.txt with pinned versions for reproducible builds.	`scripts/purevocals.py, scripts/env_manager.py:26`
低危	SKILL.md metadata lacks declared allowed-tools 文档欺骗 The SKILL.md openclaw metadata does not enumerate allowed-tools (shell:WRITE, filesystem:WRITE). While the skill does use these, the metadata block is sparse compared to the actual footprint. `requires: bins: - python` → Document allowed-tools explicitly in metadata for transparency.	`SKILL.md:1`

资源类型	声明权限	推断权限	状态	证据
命令执行	`WRITE`	`WRITE`	✓ 一致	subprocess.run/audio-separator CLI (scripts/purevocals.py), venv.create/pip inst…
文件系统	`NONE`	`WRITE`	✓ 一致	Writes to SKILL_ROOT/{models,data,logs,venv} — consistent with audio processing …
网络访问	`NONE`	`READ`	✓ 一致	PyTorch/audio-separator download URLs are legitimate ML package repos, not user-…
环境变量	`NONE`	`NONE`	—	Only reads RUNNING_IN_VENV flag to prevent venv recursion; no credential harvest…
技能调用	`NONE`	`NONE`	—	No inter-skill invocation
剪贴板	`NONE`	`NONE`	—	No clipboard access
浏览器	`NONE`	`NONE`	—	No browser access
数据库	`NONE`	`NONE`	—	No database access

3 项发现

🔗

中危外部 URL 外部 URL

https://pypi.tuna.tsinghua.edu.cn/simple

scripts/ensure_package.py:50

🔗

中危外部 URL 外部 URL

https://download.pytorch.org/whl/cpu

scripts/env_manager.py:148

🔗

中危外部 URL 外部 URL

https://download.pytorch.org/whl/

scripts/env_manager.py:161

目录结构

8 文件 · 36.9 KB · 857 行

Python 5f · 727L Markdown 2f · 90L Ignore 1f · 40L

├─ ▾ 📁 scripts

│ ├─ 🐍 config.py Python 52L · 2.1 KB

│ ├─ 🐍 ensure_package.py Python 73L · 2.8 KB

│ ├─ 🐍 env_manager.py Python 235L · 10.3 KB

│ ├─ 🐍 logger_manager.py Python 55L · 2.6 KB

│ └─ 🐍 purevocals.py Python 312L · 13.5 KB

├─ 📄 .gitignore Ignore 40L · 359 B

├─ 📝 README.md Markdown 62L · 3.5 KB

└─ 📝 SKILL.md Markdown 28L · 1.6 KB

依赖分析 6 项

包名	版本	来源	已知漏洞	备注
`pydub`	`*`	pip	否	Version not pinned
`audio-separator`	`*`	pip	否	Version not pinned
`librosa`	`*`	pip	否	Version not pinned
`ffmpeg-downloader`	`*`	pip	否	Version not pinned
`torch`	`*`	pip (PyTorch official wheel)	否	Version not pinned, GPU/CPU auto-detected
`huggingface-hub`	`*`	pip	否	Version not pinned

安全亮点

✓ All subprocess/shell calls are for legitimate audio processing tooling (audio-separator CLI, ffmpeg-downloader, venv/pip management)

✓ No credential theft, API key harvesting, or environment variable enumeration for sensitive data

✓ No data exfiltration or C2 communication detected

✓ No obfuscation, base64-encoded payloads, or anti-analysis techniques

✓ No access to ~/.ssh, ~/.aws, .env, or other sensitive paths

✓ No curl|bash or wget|sh remote script execution patterns

✓ No hidden instructions in HTML comments or prompt injection payloads

✓ All network URLs are standard PyTorch/PyPI package repositories

✓ Logs are written only to the skill's own logs/ directory