The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the google-genai Python package. This is the official, well-known SDK provided by Google for interacting with Gemini models and is considered a safe dependency.
[DATA_EXFILTRATION]: Network communication is strictly confined to official Google endpoints: generativelanguage.googleapis.com for API interactions and ai.google.dev for documentation and pricing verification. The skill handles sensitive credentials, such as the GEMINI_API_KEY, by reading them from the environment and explicitly forbidding their logging or exposure in metadata files.
[COMMAND_EXECUTION]: The documentation includes curl examples for REST API interactions and Python scripts for SDK usage. These are provided as integration guidelines and do not involve silent or unauthorized command execution. Shell operations are restricted to writing request/response JSON files within the designated generations/ workspace directory.
[PROMPT_INJECTION]: The skill processes external data by ingesting reference images and style.md files from the generations/refs/ directory to guide generation. While this presents a surface for indirect prompt injection, the risk is effectively mitigated by the skill's design, which requires explicit user approval and cost verification before each billable model interaction.

generate-nanobanana