← 返回 Skills 市场
moss-transcribe-diarize
作者
helloeveryworlds
· GitHub ↗
· v1.0.5
· MIT-0
364
总下载
0
收藏
1
当前安装
7
版本数
在 OpenClaw 中安装
/install moss-transcribe-diarize
功能描述
MOSS 多说话人转写技能。支持 URL / 本地文件 / Base64 音频输入，输出带时间戳与 speaker 的结构化转写结果（JSON、逐段文本、按说话人汇总）。用于会议纪要、访谈录音、多人对话整理。需要 API 凭证（环境变量：MOSS_API_KEY，兼容 MOSI_TTS_API_KEY / MOS...
安全使用建议
This skill appears to do exactly what it says: it uploads audio to https://studio.mosi.cn and saves structured, speaker-labelled transcripts. Before installing, verify you trust the remote service (studio.mosi.cn) because your audio and the API key are sent to that endpoint. Ensure the runtime has python3 and the 'requests' package installed (the skill does not declare a package install step). Avoid running it on highly sensitive audio unless you are comfortable with the service's privacy practices, and consider using a disposable or least-privilege API key (rotate the key later) for testing. If you need a guarantee that no external network calls occur, do not install this skill.
功能分析
Type: OpenClaw Skill
Name: moss-transcribe-diarize
Version: 1.0.5
The skill is a legitimate tool for audio transcription and speaker diarization using the MOSS API. The script `scripts/transcribe.py` correctly handles local file reading, base64 encoding, and communication with the designated endpoint (studio.mosi.cn) using provided environment variables for authentication. No evidence of data exfiltration, malicious execution, or prompt injection was found.
能力评估
Purpose & Capability
Name/description match the code and runtime instructions: the script uploads audio (URL, local file, or data URL) to a fixed transcription endpoint and returns structured, speaker-labelled output. Small implementation note: the script imports the Python 'requests' library but the skill only declares 'python3' as a required binary and provides no install spec for Python packages; this is an operational mismatch (not a security misalignment).
Instruction Scope
SKILL.md instructs the agent to run scripts/transcribe.py and the script only performs tasks required for transcription: read a local file (if provided), base64-encode it, POST JSON with the audio to the hard-coded endpoint, and write three output files. It does not read unrelated system files or other environment variables. Important privacy note: the script transmits the audio bytes and sends the API key in an Authorization header to https://studio.mosi.cn; that is expected behavior but users should be aware audio plus the key are sent off-host.
Install Mechanism
There is no install spec (instruction-only + included script). Nothing is downloaded or written by an installer. The only runtime requirement is python3 and the presence of the 'requests' Python package (not declared). No remote install URLs or archive extraction are used.
Credentials
The skill requires one of three API keys (MOSS_API_KEY as primary, or MOSI_TTS_API_KEY / MOSI_API_KEY as fallbacks). These map directly to the Authorization header used by the script and are proportional to the stated purpose. No unrelated credentials or secrets are requested.
Persistence & Privilege
The skill does not request always: true, and it does not modify other skills or system configuration. It creates output files in the working directory as expected for a transcription tool. Autonomous invocation remains enabled by default on the platform, but that is not a property of this skill specifically.
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 moss-transcribe-diarize - 安装完成后，直接呼叫该 Skill 的名称或使用
/moss-transcribe-diarize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
v1.0.5
Add user-selectable segment output formats (json/compact/text) with speaker fields; fix segment parsing for asr_transcription_result.segments.
v1.0.4
Declare credential metadata explicitly (primaryEnv + requires.env) to match runtime API-key usage.
v1.0.3
Simplify API surface to match moss-tts style: fixed endpoint, removed extra API params, keep minimal required arguments.
v1.0.2
Harden endpoint policy (HTTPS + studio.mosi.cn allowlist), clarify credential requirement, and reduce scanner false positives.
v1.0.1
Clarify required API credentials in metadata/description; align key env fallback; cleanup packaging.
v0.1.1
Improve reliability: add source validation, request timeout handling, HTTP/JSON checks, output path safety, and clearer error exits.
v0.1.0
Initial release: high-confidence diarized ASR workflow from docs, URL/file/base64 input support, structured outputs (JSON/segments/by-speaker).
元数据
常见问题
moss-transcribe-diarize 是什么？
MOSS 多说话人转写技能。支持 URL / 本地文件 / Base64 音频输入，输出带时间戳与 speaker 的结构化转写结果（JSON、逐段文本、按说话人汇总）。用于会议纪要、访谈录音、多人对话整理。需要 API 凭证（环境变量：MOSS_API_KEY，兼容 MOSI_TTS_API_KEY / MOS... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 364 次。
如何安装 moss-transcribe-diarize？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install moss-transcribe-diarize」即可一键安装，无需额外配置。
moss-transcribe-diarize 是免费的吗？
是的，moss-transcribe-diarize 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
moss-transcribe-diarize 支持哪些平台？
moss-transcribe-diarize 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 moss-transcribe-diarize？
由 helloeveryworlds（@helloeveryworlds）开发并维护，当前版本 v1.0.5。
推荐 Skills