vibevoice
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references several official Pixazo domains (
pixazo.ai,gateway.pixazo.ai,api-console.pixazo.ai) for API interaction, documentation fetching, and account management. These are vendor-owned resources and are used for their intended purpose. - [CREDENTIALS_UNSAFE]: The instructions guide the agent to manage a
PIXAZO_API_KEY. It checks for the key in environment variables and a local configuration file (~/.pixazo/api-key). It also instructs the agent to save the key to this file with restricted permissions (chmod 600). This aligns with standard secret management practices for CLI-based integration. - [COMMAND_EXECUTION]: The skill contains example code snippets in Bash (curl), Python, and Node.js. These are provided as implementation references for the agent to use when interacting with the VibeVoice API.
- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect injection as it processes user-supplied text to generate speech. However, its capabilities are confined to structured API calls to the vendor's gateway and displaying the resulting audio URL, which significantly limits the risk of instruction override.
Audit Metadata