Knowledge Card Prompt Generator

Problem

The hard part of knowledge cards is not the visual idea, but making a drawing model render exact text, keep elements in the right containers, and output multiple independent images. Plain prompts often suffer from altered wording, misplaced illustrations, and several cards merged into one canvas.

How It Works

This skill converts input text, files, or images into structured drawing prompts with explicit constraints:
- Text locking: required labels are wrapped in backticks to reduce the chance that the model rewrites them.
- Container-level placement: instructions specify where an element belongs, such as "inside the banner", "inside a circle", or "on the second pyramid layer".
- Illustration binding: each drawing cue is tied to the related text, avoiding unrelated decorative assets.
- Independent multi-image output: each card is emitted as a separate code block and marked as a full canvas.
- Aspect-ratio and splitting logic: supports 3:4 portrait and 16:9 landscape, choosing single or multiple cards based on content volume, structure, and use case.

Boundaries

It fits notes, articles, slide copy, process explanations, and comparison content for tools like Jimeng AI. Output quality still depends on how clear the source content is. If the input is dense or disorganized, it should be distilled first. If the target drawing model has prompt length limits, trim decorative wording first while preserving text, layout, and placement constraints.