The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches generated video content from external URLs in the video_generate function within exports.py. When a video generation task is completed, the skill retrieves a download_url from the Venice API response and uses requests.get to download the media file to the local workspace.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input, such as image prompts, video descriptions, and YouTube URLs, which are then used in network operations and content retrieval tasks.
Ingestion points: User-provided strings for the prompt parameter in image/video generation and external URLs provided to the image_edit, image_upscale, video_queue, and video_transcribe_youtube functions in exports.py.
Boundary markers: The instructions do not define specific boundary markers or use system instructions to isolate user-provided prompts from the agent's core logic.
Capability inventory: The skill has the ability to perform arbitrary network requests through the requests library and write files to the output/ directory on the local filesystem.
Sanitization: There is no evidence of input validation or sanitization for the URLs or prompts before they are interpolated into API requests or used to fetch content.

venice