nano-banana
Pass
Audited by Gen Agent Trust Hub on Aug 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references the installation of the
beltCLI vianpxand provides links to documentation and installation scripts hosted on GitHub (github.com/inference-sh). These are standard resources for the platform described. - [COMMAND_EXECUTION]: The skill instructions involve executing the
beltCLI tool using theBashtool. This is the primary intended function of the skill to interface with the remote image generation service. - [PROMPT_INJECTION]: The skill ingests user-provided text prompts for image generation. This represents an indirect prompt injection surface where malicious instructions could be embedded in data.
- Ingestion points: The
promptfield in the--inputJSON object within thebelt app runcommands. - Boundary markers: The skill uses structured JSON to pass arguments to the CLI, which acts as a boundary for the input data.
- Capability inventory: The skill is scoped to the
Bashtool with access restricted to thebeltcommand (belt *). - Sanitization: Input sanitization is expected to be handled by the
beltCLI and the backend inference engine.
Audit Metadata