nano-banana-2

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with ai-factory.frevana.com to send image generation requests. This is the primary function of the skill and targets the vendor's own API infrastructure.
  • [COMMAND_EXECUTION]: The script utilizes standard system tools including bash for orchestration, curl for network communication, and python3 for secure JSON payload construction and response validation.
  • [PROMPT_INJECTION]: The skill processes user-supplied text for image generation, which constitutes an indirect prompt injection surface.
  • Ingestion points: User input is captured via the --prompt or --contents arguments in SKILL.md.
  • Boundary markers: No explicit delimiters are used in the instructions to the agent to separate user input from the command context.
  • Capability inventory: The skill possesses network access (curl), script execution capabilities (bash, python3), and file system write access (via the --output flag and mkdir).
  • Sanitization: The implementation uses json.dumps within a Python helper script to ensure the user prompt is correctly escaped and safely formatted into the JSON API payload, preventing structural or command injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 03:39 PM
Security Audit — agent-trust-hub — nano-banana-2