voice-clone
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill transmits local audio samples (base64 encoded) and user-provided text to the external API domain 'api.wavespeed.ai'. This data transfer is the intended function of the skill for processing voice cloning and synthesis tasks.
- [EXTERNAL_DOWNLOADS]: The skill fetches generated audio files from the WaveSpeed AI service and saves them to the local filesystem (e.g., in '/tmp/wuna_koubo'). These downloads originate from the expected service provider domain.
- [PROMPT_INJECTION]: The skill accepts user-supplied text for synthesis, representing an indirect injection surface.
- Ingestion points: The 'text' parameter in the 'generate_speech' and 'clone_voice' functions across 'voice_clone.py' and 'SKILL.md'.
- Boundary markers: None identified; input text is interpolated directly into the JSON payload for the API.
- Capability inventory: File system read/write access and network POST/GET operations to 'api.wavespeed.ai'.
- Sanitization: No explicit sanitization or filtering is applied to the input text before it is sent to the synthesis model.
Audit Metadata