The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided slide content which is interpolated into prompts used for AI image generation. This creates a surface where embedded instructions in the content could attempt to influence the image output or downstream agents processing the deck.
Ingestion points: The --content JSON file read in scripts/generate-deck.py contains user-controlled slide prompts.
Boundary markers: Absent; user content is concatenated directly with style presets.
Capability inventory: The skill has network access via urllib.request to call the Google Imagen API and file system write capabilities to save images.
Sanitization: Absent; user input is not validated or filtered before prompt construction.
[DATA_EXPOSURE]: The script fetches AI-generated images from Google's official Generative Language API (generativelanguage.googleapis.com).
[CREDENTIALS_SAFE]: The skill uses environment variables (GEMINI_API_KEY and GOOGLE_APPLICATION_CREDENTIALS) for authentication. This is a standard and secure practice for local execution, avoiding hardcoded secrets.

deck-generator