ai-video

What This Skill Does

The ai-video skill is a powerful orchestration agent for OpenClaw that automates the generation of videos via the skills.video platform. It serves as a bridge between OpenAPI specifications and live API requests, allowing users to move from high-level creative prompts to technically sound, executed API calls without manual schema drafting. The skill handles the complexities of asynchronous video generation, specifically by leveraging Server-Sent Events (SSE) for real-time tracking, followed by intelligent polling to ensure a terminal status is returned to the user.

Installation

To install this skill, run the following command in your terminal: clawhub install openclaw/skills/skills/chuyun/gen-video

Once installed, you must configure your API key to enable authentication. Open your local OpenClaw skill settings, navigate to ai-video, and add an environment variable named SKILLS_VIDEO_API_KEY with your credentials obtained from the developer dashboard. Alternatively, you can use a shell export command: export SKILLS_VIDEO_API_KEY="your_key_here". Always run python scripts/ensure_api_key.py to verify your configuration before executing your first request.

Use Cases

• Prototyping Video Models: Rapidly test various parameters for Kling, Luma, or Sora via the skills.video platform.
• Automated Media Pipelines: Integrate video generation into larger content workflows that require programmatic output based on dynamic data.
• Debugging API Contracts: Use the inspection tools to validate that your local payloads match the requirements defined in the openapi.json specs.

Example Prompts

1. "Generate a 5-second cinematic drone shot of a snowy mountain range using the Kling v2.6 model endpoint."
2. "Inspect the OpenAPI definition for the latest video generation endpoints and create a template for a 1080p landscape animation."
3. "Check if my SKILLS_VIDEO_API_KEY is configured correctly, then prepare a request to generate a video based on my recent prompt about a futuristic cyberpunk city."

Tips & Limitations

• Always prefer SSE: The skill is optimized for SSE endpoints. Avoid falling back to polling unless the stream connection is lost or the model does not support it.
• Deterministic Payloads: Use the provided inspection scripts to generate your request templates. Guessing parameters often leads to 400 Bad Request errors due to strict schema enforcement.
• Asynchronous Handling: Remember that video generation takes time. This skill is designed to return only terminal status updates, so do not expect instantaneous completion.