senseaudio-let-claw-talkv1

当用户希望把 AudioClaw 变成一个持续监听、开口就说、停顿就回答的本机语音助手时使用。这个 skill 支持 macOS 和 Windows 两个平台：优先尝试 Python 录音链路，macOS 上再提供原生 Swift 录音兜底；用户语音通过 SenseAudio ASR 转文字，再发给 audioc...

This skill will continuously access your microphone and send recorded audio to SenseAudio's cloud endpoints (https://api.senseaudio.cn) using the SENSEAUDIO_API_KEY you provide — if you have privacy concerns, don't supply that key or don't run the skill. It stores configuration, logs, and optional model state under ~/.audioclaw/workspace (or the equivalent Windows path); review or back up those files if you care about local data. The WeSpeaker capability runs a local service and may require heavy ML dependencies (torch, torchaudio) if enabled — only enable it if you want local voice-print/model functionality. The code is self-contained (no remote installers), but you should: (1) confirm you trust the SenseAudio service before providing an API key; (2) prefer putting the key in the documented user-level credentials file rather than embedding it in the skill; (3) inspect and run the scripts in an environment you control (e.g., a user-level Python venv) if you want to limit side effects; and (4) be aware the skill will create/modify files and may launch background processes in your user workspace.

Type: OpenClaw Skill Name: senseaudio-let-claw-talkv1 Version: 1.0.1 This skill bundle implements a continuous voice assistant for macOS and Windows using SenseAudio ASR/TTS services. It includes a local speaker verification service (WeSpeaker) and cross-platform launchers that open new terminal windows to host the assistant process. While it manages API credentials in local files (senseaudio_credentials.json) and runs an unauthenticated local HTTP service for voiceprint processing, these behaviors are consistent with the stated functionality and lack evidence of malicious intent or unauthorized data exfiltration. The code is well-structured, documented, and uses standard libraries for its operations.