The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied text prompts and media (images/videos) to generate video assets. This creates a functional surface for indirect prompt injection, where malicious instructions could be embedded in user requests or media metadata to influence the agent's behavior. However, the agent's capabilities are scoped to specific video generation tools, mitigating the risk.
Ingestion points: User-provided text for motion planning and uploaded local media (first frame, tail frame, identity/product references) specified in SKILL.md.
Boundary markers: The skill explicitly instructs the agent to follow core OAuth, billing, and safety rules defined in the base Kling skill (../kling-ai/SKILL.md), though it does not define specific prompt delimiters for user input within these files.
Capability inventory: The skill utilizes specific tools including text_to_video, image_to_video, motion_control, and query_tasks to generate and monitor video tasks.
Sanitization: There is no explicit evidence of input sanitization or validation for user-supplied strings or media metadata in the provided reference files; however, the skill relies on the underlying platform's safety guardrails.

kling-ai-generate-video