bright-paper-collagelisted

# Bright Paper Collage Design Skill Turn each information item into exactly one generation prompt for the asset the caller actually requested. Apply the same recognizable collage language across still images, video, code-authored layouts, and custom media without forcing one ratio, palette, subject type, or motion pattern. ## Non-negotiable output rule - Return one `generation_prompt` per ordered item. - Compile only for the resolved `generation_mode`. - Never add a still prompt to a video request, a motion prompt to an image request, or any other companion prompt. - Never invoke ImageGen, Grok, FFmpeg, a renderer, or another asset generator. - Do not expose the internal recipe unless `include_recipe: true`. ## Accepted request Accept natural language or an ordered object containing `items`. Each item may contain: - `id`: optional stable identifier. - `information`: required source information and the only authority for factual claims. - `asset_spec`: optional `asset_type`, `generation_mode`, `aspect_ratio`, `duration_seconds`, `delivery_context`, generator, medium, dimensions, timing, and delivery constraints. - `exact_text`: optional display copy to preserve verbatim. - `text_policy`: `auto`, `none`, `preserve-reference`, or an explicit caller-defined policy. - `reference_paths`: local references whose factual and visual identity must be preserved. - `must_include`, `must_avoid`, and `design_overrides`. The supported generation modes are `text-to-image`, `image-to-im