gemini-nano-banana
Pass
Audited by Gen Agent Trust Hub on Mar 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill specifies dependencies from the author's NPM (@aiagenta2z/onekey-gateway) and PyPI (ai-agent-marketplace) packages. These are legitimate vendor resources required for the skill to communicate with the OneKey Gateway service.- [DATA_EXFILTRATION]: The skill utilizes the DEEPNLP_ONEKEY_ROUTER_ACCESS environment variable for authentication. This is a standard and secure method for handling API keys, and the scripts correctly access this variable only to pass it to the vendor's own router library.- [PROMPT_INJECTION]: The skill processes untrusted user data via image generation prompts. While this presents a surface for indirect prompt injection, the risk is negligible as it is restricted to the content of the generated image.
- Ingestion points: 'prompt' parameter in scripts/generate_image_gemini.py and scripts/generate_image_nano_banana.py.
- Boundary markers: None.
- Capability inventory: External API calls via OneKeyAgentRouter.invoke.
- Sanitization: None.- [CREDENTIALS_UNSAFE]: The Python scripts contain a fallback string 'BETA_TEST_KEY_MARCH_2026' used when the environment variable is not set. This is documented as a demo key for testing purposes and does not expose sensitive production credentials.
Audit Metadata