nano-banana
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to interact with the
beltcommand-line interface, performing actions such as authentication and executing image generation applications as documented inSKILL.md. - [EXTERNAL_DOWNLOADS]: Setup instructions in
SKILL.mdrecommend installing thebelt-sh/cliand related components vianpx skills add, which downloads code from the service provider's official repositories. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied natural language prompts which are interpolated into
beltCLI commands, creating an inherent surface for indirect prompt injection. - Ingestion points: User-provided prompt text and image URLs used in the
--inputparameter throughoutSKILL.mdexamples. - Boundary markers: Input parameters are encapsulated within JSON objects or single-quoted strings in the shell commands.
- Capability inventory: Shell command execution via the
Bashtool, restricted to thebeltexecutable by the skill's frontmatter. - Sanitization: No explicit sanitization or input filtering is implemented within the skill instructions, relying on the platform's standard command handling.
Audit Metadata