MeetingClaw - Intelligent Meeting Minutes Assistant

智能会议纪要生成工具，支持火山引擎录音文件识别2.0，自动生成结构化会议纪要。使用场景：(1) 上传音频文件生成会议纪要；(2) 语音转文字并智能总结；(3) 提取待办事项和核心观点；(4) 支持2小时以内音频处理。

Security Scan

The skill mostly matches its stated purpose (transcribe via Volcengine and summarize with a model) but contains inconsistencies in metadata and documentation vs. code that you should verify before installing.

Purpose & Capability

The skill's name/description (meeting minutes from audio using 火山引擎) align with the code: it sends audio to Volcengine and calls a large model to generate minutes. However the registry metadata states "Required env vars: none" while SKILL.md and the script require VOLC_ACCESS_KEY and VOLC_SECRET_KEY — this is an important mismatch.

Instruction Scope

SKILL.md and the script instruct the agent to: accept an uploaded audio file, save it under a workspace path, send audio binary to Volcengine async transcribe API, then send the transcript to an OpenClaw model. SKILL.md claims audio and data are only sent to local and official servers, which is consistent with using Volcengine and OpenClaw, but SKILL.md also says the local path is '~/workspace/MeetingClaw/YYYYMMDD/' while the code writes to '~/.openclaw/workspace/MeetingClaw/YYYYMMDD' — a documentation/code mismatch. The standard (TOS) flow is documented but not implemented in code (standard mode raises '暂未实现').

Install Mechanism

No install spec (instruction-only + included script). That minimizes installer risk. The script imports 'volcengine' and 'openclaw' Python modules but no install instructions are provided — confirm those dependencies exist in runtime.

Credentials

The script legitimately requires VOLC_ACCESS_KEY and VOLC_SECRET_KEY to call Volcengine, which matches the purpose. The registry metadata incorrectly lists no required env vars; this discrepancy is suspicious and should be resolved. No other unrelated secrets are requested.

Persistence & Privilege

The skill does persist files locally under a user workspace (~/.openclaw/workspace/MeetingClaw) and has a 30-day retention/cleanup. always is false and the skill does not request elevated privileges or modify other skills. Writing into the user's home directory is expected but you should verify the path is acceptable for your environment.

What to consider before installing

Key things to check before installing: - Metadata mismatch: the skill metadata claims no required env vars, but SKILL.md and scripts require VOLC_ACCESS_KEY and VOLC_SECRET_KEY. Confirm the registry entry is accurate and the publisher can be trusted. - Data flows: audio is uploaded to Volcengine (official service) and transcripts are sent to the OpenClaw model API — if meetings contain sensitive data, ensure you are comfortable with those external services and their privacy terms. - Path & retention: the code saves audio and minutes under ~/.openclaw/workspace/MeetingClaw/YYYYMMDD and automatically deletes files older than 30 days. SKILL.md documents a different path; confirm where files will actually be stored and that retention meets your policy. - Dependencies: the script imports 'volcengine' and 'openclaw' but provides no install instructions; ensure these packages are installed from trusted sources and inspect them if needed. - Unimplemented standard mode: the 'standard' (TOS upload) flow is described in docs but not implemented in the script — treat that as incomplete functionality, not a feature. - Source trust: source/homepage unknown. Prefer skills with an identified publisher or open-source repo you can audit. If you decide to proceed, supply only credentials scoped to the minimal permissions required (audio transcription and TOS access if/when used), and consider testing with non-sensitive audio first.

MeetingClaw 智能会议纪要工具

功能说明

• 支持火山引擎两种识别模式：
  1. 极速版：无需上传文件，直接提交本地音频二进制流，识别速度快，价格4.5元/小时，赠送20小时免费额度
  2. 标准版：通过火山引擎对象存储TOS上传文件生成公网URL调用，价格更低约2.5元/小时，适合长音频批量处理
• 自动生成结构化会议纪要，包含：
  1. 会议主题：智能推断核心主题
  2. 基本信息：自动读取音频创建时间，无需手动输入
  3. 会议综述：高度抽象提炼会议核心内容
  4. 会议智能总结：深度分析提炼关键观点、结论、核心问题，重要内容引用原文原话
  5. 待办事项：自动提取会后待执行任务，明确指派发言人
  6. 其他重要事项：补充遗漏关键信息
• 音频时长限制：最长2小时，与火山引擎官方限制一致
• 自动保存原始录音文件和生成的纪要，按日期归档

使用流程

1. 接收用户上传的音频文件（支持mp3/wav/m4a等常见格式）
2. 自动识别音频时长，超过2小时提示用户截断
3. 选择识别模式（默认极速版，用户可指定标准版降低成本）
4. 调用火山引擎录音文件识别接口获取转写文本
5. 调用OpenClaw大模型对转写文本进行智能分析，生成结构化纪要
6. 按日期归档保存：原始音频存放在~/workspace/MeetingClaw/YYYYMMDD/，纪要保存为同名.md文件
7. 返回生成的会议纪要给用户

配置要求

• 火山引擎API密钥：需配置VOLC_ACCESS_KEY和VOLC_SECRET_KEY环境变量
• 极速版开通：需开通「Doubao-录音文件识别2.0极速版」服务
• 标准版开通：需同时开通火山引擎对象存储TOS服务和「录音文件识别2.0标准版」服务

安全规则

1. 所有音频文件仅在本地和火山引擎官方服务器传输，不向第三方泄露
2. 本地保存的音频文件默认保留30天，到期自动清理，用户可自定义保留周期
3. 生成的会议纪要仅保存在用户本地工作区，不会上传到任何公共服务器
4. 严格遵循火山引擎数据隐私协议，转写过程数据仅用于识别，不会被留存或用于其他用途