ai-voice-cloning

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill is a legitimate voice-generation guide, but its header tells the user to install an unvetted package from a different org (belt-sh/cli) than the inference.sh sources it otherwise uses, which is sketchy without clear malicious intent.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (medium risk: 0.30). The decisive boundary is the setup instruction at line 7, which runs npx skills add belt-sh/cli to install skill material from a GitHub-style repository whose owner is not shown to be the first-party inference.sh publisher, so it is a common/reputable unofficial dependency (medium); the raw.githubusercontent.com link at line 17 is documentation for a human and, even if followed, belongs to the inference-sh organization, which would be strong.