fun-voice-type

一个语音输入法插件。它基于阿里云FunASR实时语音识别技术，允许用户通过长按快捷键（Right Option键）直接将语音转换为文字并“打”在当前光标所在的任何输入框中。此外，还能将语音翻译为多种语言（例：中英日韩）。

What to consider before installing: - The code will send microphone audio (via FunASR) and recognized text to DashScope cloud services, and may send recognized text to the qwen-plus model for translation. Do not use it for sensitive audio/text unless you trust DashScope and your API key. - You must set DASHSCOPE_API_KEY in your environment for full functionality, but that env var is not declared in the skill registry metadata — verify this discrepancy with the publisher before providing a real API key. - The script requires Accessibility/Input Monitoring permission for the terminal you run it from and will simulate typing into whatever input has focus. Granting those permissions to a terminal is powerful: consider running in a dedicated account or VM, and avoid using the tool while focused on password fields, banking apps, or other sensitive inputs. - The PKG is small and readable; if you can, review the included script yourself (or have someone you trust do so). Confirm network destinations (dashscope endpoints) and consider monitoring outbound network traffic the first time you run it. - If you need higher assurance: ask the publisher for a verified source/homepage, or request that the registry metadata be updated to declare DASHSCOPE_API_KEY as a required env var and to provide a signed release or provenance information.

Type: OpenClaw Skill Name: fun-voice-type Version: 2.0.0 The skill is a legitimate voice-to-text utility that uses Alibaba's FunASR and Qwen LLM for transcription and translation. It requires standard macOS permissions (Microphone, Accessibility) to capture audio and simulate keyboard input, which are necessary for its core functionality. The code in fun-voice-type.py is transparent, uses environment variables for API keys, and lacks any indicators of data exfiltration or malicious execution.