nano-banana-image
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill incorporates untrusted user input into prompts sent to an external image generation model, creating a surface for indirect prompt injection.\n
- Ingestion points: User inputs such as
subject,mood, andstyledefined inSKILL.mdare concatenated into the final prompt sent to the model.\n - Boundary markers: The skill lacks explicit delimiters or boundary instructions to isolate user-provided text from the system-defined prompt layers.\n
- Capability inventory: The skill utilizes the
openrouter_apito facilitate network requests to the external image model.\n - Sanitization: There is no evidence of input filtering or validation logic to prevent malicious instructions from being included in the user-provided fields.
Audit Metadata