The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes natural language prompts for image generation, which represents an ingestion surface for potentially malicious instructions embedded in data processed by the agent.
Ingestion points: The prompt parameter in scripts/generate_image.py (CLI argument) accepts arbitrary text strings that are passed to the model.
Boundary markers: There are no explicit delimiters or warnings instructing the underlying model to ignore instructions within the prompt text.
Capability inventory: The skill can perform network requests to the ModelScope API and write files to the local disk (images and configuration) via scripts/generate_image.py.
Sanitization: No input validation or sanitization is performed on the prompt content before it is encoded into the API request.
[EXTERNAL_DOWNLOADS]: The skill downloads generated image files from the ModelScope inference service after successful processing.
Evidence: The script scripts/generate_image.py fetches images from the URLs provided in the output_images field of the API response from api-inference.modelscope.cn.
Security Note: This is part of the core functionality and uses the standard requests and Pillow libraries to handle the data safely.

modelscope-zimage-generator