The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's code examples execute the local binary /home/user/.local/bin/rebyte-auth to obtain authentication credentials. This is a legitimate operation for establishing a session with the vendor's API.
[CREDENTIALS_UNSAFE]: The Python and Node.js examples read from /home/user/.rebyte.ai/auth.json to retrieve environment-specific configuration such as the API base URL. These files are part of the vendor's local toolchain.
[DATA_EXFILTRATION]: User-provided prompts and images are sent to the Rebyte API endpoint. This network activity is required for the image generation functionality and targets the vendor's own infrastructure.
[PROMPT_INJECTION]: The skill accepts user-supplied text prompts and image data which are interpolated into API payloads. While this creates a surface for indirect prompt injection, the model provider implements safety filters to block malicious content, as noted in the documentation.

nano-banana