技能 voice-system-expert
🎙️
voice-system-expert
安全 🌐
网络访问⚙️
外部命令
使用瀏覽器 AEC 實作語音系統
語音介面實作需要謹慎的架構決策，以防止音訊回授循環並確保可靠的回音消除。本技能提供使用 OpenAI Realtime API 和基於瀏覽器的回音消除來建立成熟語音架構的專業指導。
支持: Claude Codex Code(CC)
1
下载技能 ZIP
2
在 Claude 中上传
前往 设置 → 功能 → 技能 → 上传技能
3
开启并开始使用
测试它
正在使用“voice-system-expert”。 How do I prevent echo in my voice application?
预期结果:
- Use always-on microphone with browser echo cancellation enabled
- Configure getUserMedia with echoCancellation: true, noiseSuppression: true, autoGainControl: true
- Keep audio in browser pipeline using HTMLAudioElement for playback
- Avoid toggling microphone track or routing audio outside browser
- Trust browser AEC - it handles echo cancellation automatically
正在使用“voice-system-expert”。 What is the correct architecture for a voice AI assistant?
预期结果:
- Follow the industry-standard pattern used by ChatGPT, Zoom, and Google Meet
- Keep microphone enabled throughout the entire conversation
- Route audio through browser WebRTC stack to OpenAI Realtime API
- Play responses via HTMLAudioElement to maintain browser audio pipeline
- Use server-side VAD for turn detection, not microphone toggling
安全审计
安全v5 • 1/16/2026
Pure documentation skill containing only architectural guidance for voice system implementation. No executable code, scripts, network calls, or filesystem access. All 50 static findings are false positives caused by markdown documentation being misidentified as security-sensitive patterns.
2
已扫描文件
495
分析行数
2
发现项
5
审计总数
风险因素
🌐 网络访问 (1)
⚙️ 外部命令 (29)
SKILL.md:17 SKILL.md:25-38 SKILL.md:38-43 SKILL.md:43-61 SKILL.md:61-83 SKILL.md:83-92 SKILL.md:92-104 SKILL.md:104-108 SKILL.md:108-119 SKILL.md:119-126 SKILL.md:126-138 SKILL.md:138-144 SKILL.md:144-156 SKILL.md:156-167 SKILL.md:167-171 SKILL.md:171-184 SKILL.md:184-188 SKILL.md:188-198 SKILL.md:198-202 SKILL.md:202-213 SKILL.md:213-217 SKILL.md:217-220 SKILL.md:220-221 SKILL.md:221-222 SKILL.md:222-290 SKILL.md:290-291 SKILL.md:291-292 SKILL.md:292-302 SKILL.md:302-306
审计者: claude 查看审计历史 →
质量评分
38
架构
100
可维护性
87
内容
20
社区
100
安全
91
规范符合性
你能构建什么
整合語音 AI 功能
使用 OpenAI Realtime API 和 WebRTC 音訊串流為 AI 應用程式新增對話式語音功能
建立音訊管線
使用瀏覽器回音消除和適當的媒體約束建立可靠的音訊擷取和播放系統
設計語音架構
建立可防止回音回授並在所有主流瀏覽器中正常運作的語音系統模式
试试这些提示
基本語音設定
How do I set up a basic voice interface using OpenAI Realtime API with browser echo cancellation?
音訊約束
What audio constraints should I use for getUserMedia to enable proper echo cancellation?
修復回音問題
My voice system has echo problems. How do I implement the correct browser AEC pattern?
架構審查
Review my voice system code and ensure it follows the always-on microphone + browser AEC pattern without common anti-patterns
最佳实践
- Always use browser echo cancellation with echoCancellation: true in getUserMedia constraints
- Keep microphone enabled throughout the entire conversation - never toggle track state
- Route all audio through browser pipeline using HTMLAudioElement to maintain AEC
避免
- Toggling microphone track on and off during conversation breaks AEC
- Routing audio outside browser via AudioWorklet bypass breaks echo cancellation
- Implementing custom echo cancellation reinvents the wheel and causes issues
常见问题
Which browsers support browser echo cancellation?
All major browsers support WebRTC echo cancellation including Chrome, Safari, Firefox, and Edge
What is the maximum audio sample rate for voice?
Voice applications typically use 24000 Hz sample rate for optimal quality and performance
How do I handle microphone permission errors?
Catch NotAllowedError for permission denials and NotFoundError when no microphone is available
Does this skill store audio data?
No, this skill only provides guidance. Actual audio handling depends on your implementation
Why is my voice system producing echo?
Echo usually occurs when audio leaves the browser pipeline or echoCancellation is disabled
How is this different from other voice implementations?
This follows the industry-standard pattern used by ChatGPT, Zoom, and Google Meet with always-on mic