baoyu-xhs-images

Provide your source content and invoke the skill; it analyzes the content, recommends a style, layout, and image count, then asks for confirmation before generating. Each card's prompt is saved to a prompts/ directory before any image backend is called.

The skill offers 12 styles including cute, notion, chalkboard, and screen-print, plus 8 layouts such as sparse, dense, list, comparison, flow, mindmap, and quadrant. You can also use 26 presets like knowledge-card or tutorial that bundle style and layout together.

It resolves backends in priority order: a runtime-native tool such as Codex imagegen, then codex-imagegen via the codex CLI, then other native or installed backends like baoyu-image-gen. You can pin a backend with preferred_image_backend in EXTEND.md.

It generates the cover image first without references, then passes that image as a reference to every subsequent card. This image-1 anchor chain keeps character design, illustration style, and color rendering consistent across the series.

Yes, pass --yes or explicitly say to skip confirmation in your request. The skill then uses EXTEND.md preferences or built-in defaults and proceeds with the recommended plan without prompting.

The skill prohibits programmatic overlays like ImageMagick or Pillow to patch rendered text on bitmaps. If text is wrong, it regenerates from a corrected prompt file so the flawed candidate is preserved for comparison.