The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text prompts and external media assets (images, audio) via URLs to generate video, creating a potential surface for indirect prompt injection. * Ingestion points: User-supplied prompt and external media URLs (image, last_frame_image, audio). * Boundary markers: The skill provides instructions for prompt crafting but does not specify delimiters or sandboxing instructions for the final API payload. * Capability inventory: The skill performs network operations via curl to the Pruna API. * Sanitization: No sanitization or escaping of user-provided content is described in the skill instructions.
[EXTERNAL_DOWNLOADS]: The documentation provides commands to install prerequisite and related skills using npx skills add PrunaAI/pruna-skills@<skill-name> -y, which downloads code from the vendor's repository.
[COMMAND_EXECUTION]: The skill provides functional curl examples for interacting with the api.pruna.ai infrastructure for file uploads and prediction requests.
[PROMPT_INJECTION]: The skill documentation lists disable_safety_filter as a common optional field for the API call, which represents a capability to bypass safety constraints at the model level.

p-video