← 返回 Skills 市场
139
总下载
0
收藏
1
当前安装
2
版本数
在 OpenClaw 中安装
/install audio-command-executor
功能描述
Processes inbound audio files, transcribes them, and answers to resulting texts. Converts non-WAV inputs to WAV before transcription.
使用说明 (SKILL.md)
Body
Trigger
- Inbound Audiodateien, die im Verzeichnis /home/sirko/.openclaw/media/inbound/ landen (z. B. .ogg, .mp3, etc.)
Input
- Eingabe: Pfad zur Audiodatei (z. B. /home/sirko/.openclaw/media/inbound/aufnahme.ogg)
Workflow
- Normalize Format
- Wenn Input nicht .wav ist, konvertiere zu WAV: /usr/bin/ffmpeg -i {input_file} -ar 16000 -ac 1 -c:a pcm_s16le {input_file}.wav Hinweis: Die Zieldatei heißt input_file.wav (Beispiel: /.../aufnahme.ogg → /.../aufnahme.wav)
- Transkription
- Transkribiere die WAV-Datei: /home/sirko/.openclaw/workspace/whisper.cpp/build/bin/whisper-cli -l DE -np -m /home/sirko/.openclaw/workspace/whisper.cpp/models/ggml-small.bin -f {input_wav_file}
- Fange die Transkription als Text ab (stdout)
- Ausführung
- aus dem transkribierten Text entstandene Fragen oder Anweisungen einfach so in deutsch beantworten, als wäre es ein normaler Text, eingegeben über den Chat
Output
- Einfach den Text verarbeiten, als wäre er als Text-DM eingegangen
- Bei Fehlern: klare Fehlermeldung mit Ursachen (z. B. Datei nicht gefunden, Transkript leer, Ausführung fehlschlägt)
Beispiel-Ablauf
- inbound/file.ogg → convert → /tmp/file.wav → whisper → "Was ist die Hauptstadt von Frankreich" → ermittele Antwort und zeige sie
Notes
- immer auf deutsch antworten
Tests/Testszenarien
- Test mit file.ogg (4 Sekunden) → Transkription prüfen
- Test mit bereits WAV-Datei → direkte Transkription
- Test mit fehlerhafter Datei → ordentliche Fehlermeldung
安全使用建议
Before installing, consider: (1) The SKILL.md hardcodes user-specific paths (/home/sirko/...) and absolute binary paths (/usr/bin/ffmpeg, a local whisper-cli). Verify and edit these paths to match your environment so the skill won't try to access another user's directories. (2) Ensure ffmpeg, whisper-cli, and the referenced model file actually exist and are trusted — the skill will execute them. (3) Test with non-sensitive audio first, since transcription may expose private content to the agent. (4) Ask the publisher for corrected metadata: list required binaries and any model files, and remove or parameterize hardcoded paths. If you cannot verify or adjust these issues, avoid installing or run it in a restricted/test environment.
功能分析
Type: OpenClaw Skill
Name: audio-command-executor
Version: 1.0.1
The skill exhibits critical vulnerabilities in SKILL.md, specifically shell command injection risks where the {input_file} variable is passed unsanitized to ffmpeg and whisper-cli. Furthermore, the instructions explicitly direct the agent to treat transcribed audio content as direct commands, creating a high risk for indirect prompt injection. While these appear to be design flaws rather than intentional malware, the combination of shell execution and hardcoded absolute paths (/home/sirko/) poses a significant security risk.
能力评估
Purpose & Capability
The skill claims to process inbound audio and transcribe it, which aligns with the instructions. However the SKILL.md expects /usr/bin/ffmpeg and a local whisper-cli binary at /home/sirko/.openclaw/workspace/whisper.cpp/... and references inbound files under /home/sirko/.openclaw/media/inbound/. The registry metadata declared no required binaries or environment variables — that is inconsistent with the runtime commands the skill instructs the agent to run.
Instruction Scope
Runtime instructions are narrowly scoped to converting an input audio file and transcribing it, then responding in German. But they reference hardcoded absolute paths tied to a specific user (/home/sirko/...) and call local binaries by absolute path. This makes the skill environment-specific and could cause it to fail or behave unexpectedly on other systems; it also means the agent will read files from that user path and execute local binaries.
Install Mechanism
There is no install spec (instruction-only). The included scripts/package_skill.py is a simple packaging utility with no network calls or hidden behavior. Lack of an install step reduces risk of arbitrary code being pulled during install, but runtime still executes local binaries.
Credentials
The skill declares no environment variables or credentials, which is reasonable for a local transcription skill. However, the instructions implicitly require filesystem access to /home/sirko/... and presence of a local model file and whisper binary; the absence of declared required binaries (ffmpeg, whisper-cli) is an omission that could mask necessary privileges or assumptions about the host environment.
Persistence & Privilege
always is false and the skill is user-invocable. The skill does not request persistent privileges or attempt to modify other skills or global agent settings in the provided files.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install audio-command-executor - 安装完成后，直接呼叫该 Skill 的名称或使用
/audio-command-executor触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.1
- Updated ffmpeg command in documentation to reference /usr/bin/ffmpeg for audio conversion.
- Added scripts/package_skill.py to the repository.
- Bumped version from 1.0.2 to 1.0.3.
v1.0.0
audio-command-executor 1.0.2
- Added automatic conversion of non-WAV audio files (e.g., .ogg, .mp3) to WAV format before transcription.
- Improved error handling with clearer messages for issues like missing files or empty transcriptions.
- Now processes transcribed audio commands as if they were normal chat text, always responding in German.
- Updated documentation with a detailed workflow, example usage, and specific test cases.
元数据
常见问题
Audio Command Executor 是什么？
Processes inbound audio files, transcribes them, and answers to resulting texts. Converts non-WAV inputs to WAV before transcription. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 139 次。
如何安装 Audio Command Executor？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install audio-command-executor」即可一键安装，无需额外配置。
Audio Command Executor 是免费的吗？
是的，Audio Command Executor 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Audio Command Executor 支持哪些平台？
Audio Command Executor 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Audio Command Executor？
由 Sirko（@sirkovz）开发并维护，当前版本 v1.0.1。
推荐 Skills