The Agent Skills Directory

[SAFE]: The skill's primary function is to act as a documentation and routing layer for Higgsfield video generation tools. It does not contain any executable scripts, shell commands, or network exfiltration patterns.
[EXTERNAL_DOWNLOADS]: The skill references multiple external documentation sources and vendor platforms including Higgsfield (github.com/higgsfield-ai, higgsfield.ai), Google (ai.google.dev, cloud.google.com), xAI (docs.x.ai), Kling (app.klingai.com), ByteDance (docs.byteplus.com), and Alibaba (help.aliyun.com). These references are neutral, serving to provide the agent with official prompt-crafting guidelines, and do not involve the execution of remote code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user input (descriptions of subjects, scenes, and actions) and incorporate them into prompts for video generation tools.
Ingestion points: SKILL.md defines a workflow to collect user slots for scene and camera directions.
Boundary markers: The instructions do not specify explicit delimiters for the final tool call, but emphasize strict adherence to vendor-specific prompt frameworks.
Capability inventory: The skill is limited to calling specified MCP tools for catalog exploration and video generation (models_explore, generate_video, etc.).
Sanitization: While no manual sanitization is described, the skill relies on the underlying tool's schema validation and the model's own safety guardrails.

media-higgsfield-video