video-prompting

name	video-prompting
description	Draft, refine, or use model-ready prompts for AI Media Canvas video generation. Use when the user asks for video prompts, prompt optimization, storyboard-to-video prompts, image-to-video motion prompts, keyframe/reference-image prompts, or finished video generation with supported models such as Seedance, Veo, Kling, Sora, Wan, Hailuo, Runway, Agnes, Grok Imagine, or HappyHorse.

Video Prompting

Create production-ready prompts for AIMC's generate_video tool or for users to paste into external video models. This is a prompt-director skill, not a separate provider or API path.

Workflow

1. Identify the target model family from the user's words or selected model id. If absent, default finished AIMC generation to google-official/veo-3.1-generate-preview; default prompt-only drafting to a generic cinematic prompt and mention the best-fit model family only if useful.
2. Identify input mode: text, image, keyframes, reference, multivideo, or video. For image/video inputs, label each asset role before drafting.
3. Read references/model-formats.md when model-specific constraints, aliases, or format details matter.
4. Draft the prompt around visible action: subject, setting, motion beats, camera, lighting/style, and audio.
5. Keep generation controls outside the prompt unless the model style benefits from time beats. Return or call with separate parameters: model, duration, resolution, aspectRatio, videoMode, inputImages, inputVideo, enableAudio, negativePrompt, seed.

Default Output