openclaw-whisper-asr
v1.0.0本地 Whisper 语音识别配置。自动将飞书/Telegram 等渠道的语音消息转成文字。 适用于需要离线、低延迟语音转文字的场景。
⭐ 0· 191·1 current·1 all-time
byxiaohuozi@279458179
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (local Whisper ASR for voice messages) aligns with the instructions: cloning whisper.cpp, downloading models, installing ffmpeg, converting audio and running whisper-cli. All required actions are coherent with offline speech-to-text.
Instruction Scope
SKILL.md instructs performing system operations (apt-get, git clone, running a model download script, cmake build, ffmpeg conversions). These are expected for building/running whisper.cpp, but they grant the skill operator the ability to run arbitrary build scripts and write into /home/brew/.openclaw/workspace — verify you are comfortable with those file writes and that commands are run as a non-privileged user.
Install Mechanism
No formal install spec (instruction-only). The guide clones from GitHub and runs the project-provided model download script and build steps. Cloning and running upstream scripts is normal here but carries the usual risk of executing third-party code — check the repository and downloaded scripts before running.
Credentials
The skill does not request any environment variables, credentials, or config paths beyond writing to a workspace path. No unnecessary secrets are asked for.
Persistence & Privilege
always is false and the skill does not request persistent elevated privileges or modify other skills' configs. It recommends creating files under a workspace path; that is typical for a local installation.
Assessment
This is a how-to for running whisper.cpp locally and is coherent with its stated purpose, but before following it you should: 1) review the GitHub repo (https://github.com/ggml-org/whisper.cpp) and any download scripts to ensure you trust the source; 2) run build steps as an unprivileged user (avoid running arbitrary build scripts as root); 3) be aware the model download will pull large binaries from the network — verify URLs and checksums if available; 4) confirm the workspace path (/home/brew/.openclaw/workspace) is appropriate for your environment or change it to a directory you control; 5) consider using a container or VM if you prefer isolation. If you need confirmation about specific download URLs or want guidance to run this inside a container, provide details and I can help.Like a lobster shell, security has layers — review code before you run it.
latestvk978gt42264ds2f47m6h0emb4s8340zt
License
MIT-0
Free to use, modify, and redistribute. No attribution required.