nano-banana-2
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a legitimate integration for the inference.sh service to perform image generation and editing. No malicious patterns such as obfuscation, credential exfiltration, or unauthorized persistence were identified.
- [COMMAND_EXECUTION]: The skill documentation includes examples of using the
beltCLI tool. The environment is properly restricted to this command via theallowed-toolsconfiguration in the frontmatter, limiting the agent's shell access. - [EXTERNAL_DOWNLOADS]: The skill references installation instructions for the
belttool hosted on GitHub and mentions theinferenceshPython library, both of which are legitimate dependencies for the service.
Audit Metadata