飞书语音消息发送技能。使用 Edge TTS（微软，免费无限次）生成语音并以飞书语音气泡形式发送。当用户要求用语音回复、发语音消息、TTS 朗读内容时触发。默认音色：云希（zh-CN-YunxiNeural，活泼阳光男声）。

Before installing or running this skill: (1) Inspect ~/.openclaw/openclaw.json and confirm it contains only the Feishu appId/appSecret you intend to use — don't point the script at a file containing broad secrets. (2) Prefer creating a Feishu app with minimal permissions for this purpose. (3) Because the registry metadata does not declare the required config/credentials, treat that omission as a red flag: either ask the publisher to update the metadata or avoid installing until it's fixed. (4) Review the included scripts/send_voice.py yourself (it is small and readable); if you run it, do so in an environment where executing curl/ffmpeg/edge-tts is acceptable (or sandboxed). (5) If you want stronger guarantees, request the skill be updated to accept credentials via a clearly-declared config path or dedicated environment variables (and update registry metadata accordingly). Additional info that would raise confidence to 'high': metadata that explicitly declares the required config path and primary credential, or an updated script that only reads a minimal, well-documented credentials file.

Type: OpenClaw Skill Name: feishu-edge-tts-voice Version: 1.0.0 The skill bundle is designed to send voice messages via Feishu using Edge TTS and FFmpeg. All network calls are directed to the legitimate Feishu API (`open.feishu.cn`). While the script `scripts/send_voice.py` uses `subprocess.run` for `ffmpeg` and `curl`, the inputs to these commands (temporary file paths, API tokens from JSON responses, hardcoded parameters) appear to be controlled and do not present clear shell injection vulnerabilities. The `SKILL.md` instructions are straightforward and do not contain any prompt injection attempts against the agent. The script accesses Feishu credentials (`appId`, `appSecret`) from `~/.openclaw/openclaw.json`, but this is necessary for its stated purpose and there is no evidence of exfiltration or misuse of these credentials.