nano-banana-image-generator
Pass
Audited by Gen Agent Trust Hub on Mar 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions require the agent to execute terminal commands such as
curlorwgetto save images locally. - [EXTERNAL_DOWNLOADS]: The skill downloads content from external S3 URLs (a well-known service) to the local directory
/Users/mac/Public/images/. - [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface where user-supplied image prompts are used to construct filenames for shell commands.
- Ingestion points: The image
promptparameter provided by the user. - Boundary markers: No delimiters or warnings are provided to prevent the agent from including shell metacharacters in the filename.
- Capability inventory: The workflow utilizes shell command execution (
curl/wget) and file system write access. - Sanitization: The skill lacks instructions to escape or sanitize characters in the kebab-case filename derived from user input, which could be exploited for command injection.
Audit Metadata