voice-cloning

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a standard workflow for utilizing voice cloning models from Alibaba and Runware. No malicious code, obfuscation, or unauthorized access patterns were detected.
  • [DATA_EXPOSURE]: The skill processes user-provided audio recordings and text descriptions. This data ingestion is central to its stated purpose and is handled through standard platform tools. The inclusion of a "Consent / likeness" section indicates awareness of data ethics.
  • [EXTERNAL_DOWNLOADS]: References to models like 'alibaba:qwen' and 'runware:dia' are for legitimate inference purposes. These sources are considered well-known or vendor-owned and do not present a security risk.
  • [PROMPT_INJECTION]: The skill utilizes natural language inputs for voice design and synthesis. While this represents a potential surface for indirect prompt injection, the output is limited to generated audio files, which does not provide a path for system-level exploitation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 04:08 PM
Security Audit — agent-trust-hub — voice-cloning