声音克隆技能。当你需要提供一段参考音频，并生成使用该声音说话的新音频时调用此技能。

This skill appears to implement voice cloning but has several red flags you should consider before installing or using it: - The script will send files (audio/images/videos) to a remote API. The default endpoint (https://dlazy.com) and the TEAM_BASE_URL env var are not documented in the skill metadata — verify the intended service and its privacy/security properties before providing your API key or any reference audio. - SKILL.md explicitly forbids running via the platform runner and demands executing the bundled Python script directly. This bypass could be an attempt to avoid platform controls; prefer skills that operate through the platform's approved invocation methods. - TEAM_BASE_URL is used but not declared; ensure the API key you supply is scoped/rotatable and do not reuse high-privilege credentials. Consider creating a throwaway or limited-scope key for testing. - The script requires Python packages (requests) that are not installed automatically; run in an isolated environment (container/VM) if you must test it. - If you need this capability, ask the publisher for a verified homepage, a canonical base URL, and a clear justification for bypassing openclaw run. If those aren't provided, treat the skill as risky and avoid sending private or sensitive audio files to it.

Type: OpenClaw Skill Name: lrshuai-voice-clone Version: 1.0.0 The skill bundle contains instructions in SKILL.md that explicitly command the AI agent to bypass the standard 'openclaw run' execution path, which may be intended to circumvent environment restrictions or logging. The script script/invoke_model.py exfiltrates local files (encoded as Base64) and the TEAM_API_KEY to a hardcoded third-party domain (dlazy.com). Furthermore, there is a significant discrepancy between the documentation, which claims to be a voice cloning tool using an '--audio' flag, and the actual code, which lacks that flag but includes logic for uploading images and videos.