📦 Audio Enhancement Engine — 音频增强引擎
v1.0.2集成 **VoiceFixer**（通用语音修复）与 **AudioSR**（高保真音频超分辨率）
0· 22·0 当前·0 累计
by 下载技能包
最后更新
2026/4/19
安全扫描
OpenClaw
可疑
medium confidence该技能整体上类似于一款离线音频增强工具，但它会执行广泛的环境变更（创建 venv、从 PyPI/GitHub 安装包、下载大型 wheel/模型）并探测系统（nvidia-smi、find），比单纯的文件处理器更具侵入性，需谨慎对待。
评估建议
This skill appears to be what it claims (an ML-based audio enhancer), but it is invasive: on first run it will create a virtual environment, install many Python packages (including from GitHub), download model weights and possibly multi-GB PyTorch wheels, probe GPU and system library paths, and re-launch itself inside the created venv. These actions can execute remote code (pip install from git+ URLs) and modify LD_LIBRARY_PATH and the filesystem. Before installing/runing: 1) Ensure you trust th...详细分析 ▾
ℹ 用途与能力
Name/description (VoiceFixer + AudioSR local audio enhancement) matches the code: scripts implement VoiceFixer and AudioSR flows and accept file/folder inputs. Requiring only 'python' is plausible for a Python ML skill. The heavy dependency installs, venv management and GPU detection are consistent with an ML audio tool, so purpose and capability are generally aligned.
⚠ 指令范围
SKILL.md limits scope to audio files, and the core logic adheres to that. However, runtime instructions/code do more than just read/process audio: they create virtual environments, run pip installs (including git+ URLs), probe hardware (nvidia-smi), search filesystem for libraries (find across /usr/local), adjust LD_LIBRARY_PATH, and re-exec the main script in a new interpreter. These system-level actions go beyond narrow 'open file → enhance → write output' behavior and increase the attack/impact surface.
⚠ 安装机制
No declared install spec, but code auto-installs dependencies at runtime using pip (including git+github URLs). ensure_package supports installing arbitrary specs and fallback local zip files; upgrade_torch downloads large wheels from download.pytorch.org. Installing packages from git+ or running pip install executes remote code at install time — this is expected for ML model repos but is higher risk than an instruction-only skill or curated package install.
ℹ 凭证需求
The skill declares no required credentials or env vars, which matches metadata. In practice it reads/sets environment variables (RUNNING_IN_VENV, VOICEFIXER_CACHE_DIR, HF_HUB_* and modifies LD_LIBRARY_PATH) and will create a venv, logs, and models directories under the skill root. There are no secrets requested, but filesystem and environment modifications are substantial relative to a simple audio converter.
ℹ 持久化与权限
always:false and no cross-skill config changes. The skill persists artifacts (venv, downloaded models, logs, downloaded wheels) on disk under its project paths and may re-run itself in the venv. It does not request global agent privileges, but the automatic venv creation and package installs give it long-lived presence on the host filesystem.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.22026/4/19
- 技能名称从 audio-enhancement 更新为 audio-enhancement-engine。 - 其余内容无更改。
● 可疑
安装命令
点击复制官方npx clawhub@latest install audio-enhancement-engine
镜像加速npx clawhub@latest install audio-enhancement-engine --registry https://cn.longxiaskill.com