The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's test suite in tests/test_nanobanana.py utilizes subprocess.run() to execute the primary CLI script for integration testing. This is a standard development practice and is not active during production skill use.
[DATA_EXFILTRATION]: The core script scripts/nanobanana.py transmits user prompts, local image data, and API keys to Google's official Gemini API endpoints (a well-known service). This communication is the primary intended function of the skill and is documented for the user.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted data in the form of text prompts and local image files provided through CLI arguments.
Boundary markers: Untrusted data is encapsulated within a structured JSON object before being sent to the remote API.
Capability inventory: The skill possesses network access via urllib.request and file system access to read inputs and save generated images.
Sanitization: No explicit content sanitization is performed on the prompt text; the skill relies on the downstream model's safety guardrails.
[DYNAMIC_EXECUTION]: The integration tests in tests/test_nanobanana.py use the importlib module to dynamically load the subject script for validation. This usage is restricted to the test environment and follows standard Python testing patterns.

nanobanana