SOTA Zero-shot Voice Cloning TTS

Voice-first OpenClaw skill powered by MOSS APIs. Use when a user wants spoken replies in a preferred timbre, either from an existing voice_id or from a refer...

Before installing, confirm the provider and provenance: ask the publisher to (1) update registry metadata to list the required MOSI_API_KEY and MOSI_BASE_URL, (2) provide a homepage or source repo so you can verify the service and privacy policy, and (3) clarify whether 'MOSS' vs 'MOSI' is a typo or different provider. Treat the MOSI_API_KEY like any secret — only grant a key scoped to minimal permissions and avoid reusing high-privilege keys. Understand that using local audio will upload user files to the external service (privacy risk); if that is unacceptable, do not enable local-file uploads. If you proceed, consider restricting the skill's network access or using an API key with limited quota/expiration, and verify logs to ensure keys are never echoed.

Type: OpenClaw Skill Name: moss-tts Version: 1.0.0 The skill defines a workflow that involves fetching content from a user-provided `reference_audio` URL and uploading files specified by a `local audio path` to an external API (`https://studio.mosi.cn`). While these actions are necessary for the skill's stated purpose (TTS with custom voices), they introduce potential Server-Side Request Forgery (SSRF) and arbitrary file upload vulnerabilities if the AI agent's underlying execution environment does not properly sanitize or restrict these inputs. The `SKILL.md` file does not contain explicit malicious instructions, but the capabilities themselves are high-risk if exploited due to implementation flaws in the agent.