Voice Broadcast

语音播报控制技能。将AI回复内容转换为语音朗读。触发方式：（1）用户说"朗读"时，自动将AI最后一条文字回复转为语音；（2）用户说"开启语音播报"时，之后所有回复自动朗读；（3）用户说"静音"时，暂停语音播报。用于：用户（尤其是iOS用户）希望通过语音方式接收信息，或双手不便时通过TTS播放回复内容。

Security Scan

The skill's described purpose (TTS voice broadcasting) mostly matches its instructions, but there are incoherent or risky details — notably Feishu delivery is required but no credentials are declared, the skill forces broadcast of 'urgent' content (overriding mute), and it writes persistent state — these gaps merit clarification before install.

Purpose & Capability

Name/description (voice broadcast/TTS for replies) matches the instructions to parse commands, use a TTS tool, and send audio. However SKILL.md requires sending voice via Feishu (飞书) but the skill does not declare any Feishu credentials or environment variables; that omission is unexpected unless the platform itself provides a built-in Feishu tool.

Instruction Scope

Instructions read/write /workspace/memory/voice_state.json (reasonable for local state) and call a 'tts' tool then send audio via Feishu. Two problematic behavioral items: (1) the skill forces broadcasting of 'urgent' clinical/critical content and explicitly ignores mute, which overrides user preferences and can leak sensitive info; (2) it says the TTS tool will auto-send audio and set textual reply empty, which is a side-effect that may be surprising and could hide textual context or logging. The SKILL.md gives the agent broad discretion to decide 'urgent'—this is vague and risky.

Install Mechanism

Instruction-only skill with no install spec or code files — lowest install risk. Nothing is downloaded or written beyond the declared workspace state file.

Credentials

No environment variables or credentials are requested. That is reasonable if the platform provides built-in 'tts' and Feishu tools, but inconsistent if Feishu sending requires tokens: sending messages to a channel usually needs credentials (e.g., Feishu bot token) which are not declared. The skill does not request any unrelated secrets or system paths.

Persistence & Privilege

The skill writes a small persistent state file under /workspace/memory/voice_state.json to track auto/mute flags — this is proportionate for its function. always:false and normal invocation settings are appropriate. It does not request system-wide privileges or modify other skills.

What to consider before installing

Before installing, confirm the platform's environment: does it provide a built-in 'tts' tool and Feishu integration so no credentials are needed? If not, ask the author which credentials/config are required and why they weren't declared. Decide whether you accept the 'urgent content forced broadcast' behavior — this will override mute and could expose sensitive information. Verify where /workspace/memory/voice_state.json is stored and whether other skills or users can read it. Test the skill in a restricted environment first and request the author to add explicit handling for 'urgent' rules (e.g., a configurable whitelist) and to declare any required credentials.

语音播报 Skill

状态文件

• /workspace/memory/voice_state.json — 记录开关状态

  { "auto": false, "muted": false }
  

触发命令

用户输入	行为
朗读	将AI当前回复朗读出来
开启语音播报	设置 auto=true，之后每条回复自动朗读
关闭语音播报	设置 auto=false，暂停自动朗读
静音	设置 muted=true，临时静音（保持 auto 状态）
取消静音	设置 muted=false

执行流程

1. 解析用户消息中的命令关键词
2. 更新状态文件（读写 /workspace/memory/voice_state.json）
3. 朗读动作：使用 tts 工具，将AI回复文本转为语音
4. 通过飞书发送语音消息（asVoice 模式）

播报优先级

• auto=true + muted=false → 自动朗读每条回复
• auto=true + muted=true → 静默，跳过朗读
• auto=false → 仅"朗读"命令触发

TTS 使用规则

• 使用 tts 工具，channel 留空（默认音频输出）
• 文本长度超过500字时，分段朗读或取前500字精华
• 回复末尾加上 " 🤖 已朗读" 提示

状态切换

• 开启语音播报 → auto=true（每次回复自动朗读）
• 关闭语音播报 → auto=false（恢复文字模式）
• 静音 → muted=true（临时静音，不改变auto状态）
• 取消静音 → muted=false（恢复播报）

注意

• TTS 工具会自动发送音频，调用后 reply 置空
• 语音播报仅支持飞书频道，其他频道静默处理
• 危急内容（如急诊处理、危急值）强制播报，忽略静音