wtf is assetflow-skill?

AssetFlow is an automated pipeline that helps AI agents handle video files without getting confused. The README describes the core problem clearly: when AI coding agents try to work with video programmatically, they tend to hallucinate incorrect file paths, struggle with timeline calculations, and crash video renderers by feeding them clips with inconsistent frame rates or resolutions. AssetFlow solves this by giving agents a defined set of scripts they can call to generate, fetch, trim, normalize, and measure media files in a predictable way. The workflow is built around HeyGen Hyperframes, a tool for assembling programmatic video. AssetFlow acts as the layer between raw, unedited clips and the Hyperframes renderer. An agent reads a SKILLS.md file that documents what each script does, then runs those scripts as needed: listing available local assets, generating an AI avatar clip via the HeyGen API, getting the exact duration of a video file, trimming a raw recording to a specific moment, and normalizing clips to a consistent resolution and frame rate before passing them to Hyperframes. It also integrates with ElevenLabs for premium voiceover generation and Ideogram for AI-generated image backgrounds, with each service requiring its own API key set in an environment file. The tech stack is Node.js with no complex frameworks for the core scripts, FFmpeg and FFprobe for video processing, and a separate Next.js and Tailwind CSS gallery UI for browsing community assets. It is written primarily in TypeScript and is designed to be installed into an existing AI agent project via a one-line curl script.