digital-avatar-shopping-video
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill includes technical guides for using
subprocess.runto invokeffmpegfor video assembly. This is a standard and necessary function for the skill's primary purpose of synthesizing video layers (avatars, products, and UI). The implementation uses argument lists rather than raw shell strings, which minimizes command injection risks.\n- [EXTERNAL_DOWNLOADS]: The system integrates with reputable external AI services for speech-to-text and digital avatar generation, including Microsoft Azure, Baidu, HeyGen, and D-ID. These connections are used to fetch generated media assets based on user-provided scripts.\n- [CREDENTIALS_UNSAFE]: The integration scripts correctly demonstrate how to manage API keys using environment variables (e.g.,os.getenv("COZE_HEYGEN_API_KEY_...")) instead of hardcoding secrets within the skill's source code, following standard secure development practices.\n- [SAFE]: The skill defines strict persona guardrails ("小省导购员") and content restrictions to prevent the generation of deceptive or inappropriate advertising content. No evidence of persistence mechanisms, credential exfiltration, or obfuscation was found during the analysis.
Audit Metadata