The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation and parameter table specify a disable_safety_filter parameter which is set to true by default. This instruction effectively requests the underlying service to bypass standard safety protocols and content filtering during video generation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data which is then passed to an external execution tool.
Ingestion points: Data is ingested through the voice_script, voice_prompt, and video_prompt parameters in the JSON payload (SKILL.md).
Boundary markers: No boundary markers or specific instructions to treat these fields as data rather than instructions are provided.
Capability inventory: The skill executes shell commands via the belt CLI tool, which communicates with external APIs.
Sanitization: There is no evidence of sanitization or validation of the input strings before they are incorporated into the tool call.
[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions to install external dependencies and CLI tools from the service provider's official infrastructure.

p-video-avatar