The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several local Python scripts (detect-backend.py, generate_image.py, nano-banana-generate.py, nano-banana-process.py) to manage the image generation lifecycle and post-processing tasks.
[EXTERNAL_DOWNLOADS]: The workflow requires the installation of the google-genai and pillow Python packages from standard public registries.
[CREDENTIALS_UNSAFE]: The documentation includes instructions for the agent to verify the existence of the GEMINI_API_KEY by echoing its last four characters. While intended for debugging, this results in the partial exposure of a sensitive credential in the execution logs.
[PROMPT_INJECTION]: The workflow processes prompts from external files and JSON manifests, which presents a surface for indirect prompt injection.
Ingestion points: Markdown files in the prompts/ directory and JSON manifests containing image generation prompts.
Boundary markers: The skill does not employ explicit boundary markers or instructions to ignore embedded commands within the prompt data files.
Capability inventory: The agent executes local Python scripts with command-line arguments derived from the content of these files.
Sanitization: There is no evidence of sanitization or validation of the prompt content before it is interpolated into shell commands.

image-gen