The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external inputs that could potentially be used to influence the behavior of the image generation model or the agent.
Ingestion points: The skill accepts text prompts and reference image paths or URLs via command-line arguments or YAML configuration files processed by scripts/generate_image.py.
Boundary markers: No explicit delimiters or instructions are used to separate user-provided content from the systemic prompt logic sent to the API.
Capability inventory: The skill's script can perform network POST requests, read local image files from the filesystem, and write downloaded image files back to the disk.
Sanitization: There is no evidence of prompt filtering or image content validation before the data is transmitted to the external API provider.
[DATA_EXFILTRATION]: The scripts/generate_image.py script reads local image files specified by the user and uploads them to the Volcengine Ark API (ark.cn-beijing.volces.com). This behavior is consistent with the skill's primary purpose of image-to-image generation.
[EXTERNAL_DOWNLOADS]: The skill performs an HTTP GET request to download the resulting image from a URL provided by the Volcengine API response.
[COMMAND_EXECUTION]: The documentation instructs the agent or user to execute a Python script using the uv run command, which is a standard execution pattern for this environment.

generate-image-by-seedream