genmedia-image-artist
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external media files including images, videos, and PDF documents through the
nanobanana_image_generationtool, creating a surface for indirect instructions to enter the agent's context. - Ingestion points: The
images,video, andPDFparameters of the generation tool allow for the ingestion of potentially malicious external content. - Boundary markers: The instructions lack delimiters or specific directives to ignore instructions embedded within the processed data.
- Capability inventory: The skill is authorized to use the
mcp_nanobanana_nanobanana_image_generationtool to generate and output visual assets. - Sanitization: No specific sanitization or filtering logic is mentioned for handling content from the provided external files.
- [PROMPT_INJECTION]: The instructions provide a method for the agent to circumvent platform safety mechanisms.
- Evidence: The skill instructs the agent to perform a "clinical rewrite" to remove labels that trigger safety filters while retaining the original intent, which is a documented evasion technique against content safety guidelines.
Audit Metadata