Cast content into a visible form. Content goes in, PNG comes out. The mold determines the shape.
Parameters
| Parameter | Mold | Size | Description |
|---|---|---|---|
-l(default) |
Long image | 1080 x auto | Single reading card, content automatically expands in height |
-i |
Infographic | 1080 x auto | Content-driven adaptive visual layout |
-m |
Multi-card | 1080 x 1440 | Automatically splits into multiple reading cards |
-v |
Visual notes | 1080 x auto | Editorial magazine feature: problem→failure→turning point→epiphany→naming (6 layout molds / 4 typeface comparison / detective file details) |
-c |
Comic | 1080 x auto | Japanese black-and-white comic style, dynamically selects the manga artist's visual language |
-w |
Whiteboard | 1080 x auto | Whiteboard marker style, structured boxes + arrows + colored marks |
-b |
Large text | 1080 x 1440 | Stele-style large text + washi paper + outer shadow, Xiaohongshu attachment style (single sentence/short paragraph) |
Constraints
This skill outputs visual files (PNG), so the Org-mode, Denote, and ASCII-only specifications in L0 do not apply.
Shared foundation
Obtain content
- URL --> WebFetch to obtain
- Pasted text --> use directly
- File path --> Read to obtain
File naming
Extract the title or core idea from the content as {name}(use Chinese directly, remove punctuation, ≤ 20 characters).
Screenshot tool
node assets/capture.js <html> <png> <width> <height> [fullpage]
Run from the skill root directory. Depends on the node_modules/ in the skill root directory for playwright. If an error occurs:
npm install playwright && npx playwright install chromium
Footer
- Left side: logo + 李继刚 (already hardcoded in the template)
- Right side: content source (optional) — displayed when there is a clear source (such as author name, arxiv ID, website name, etc.), left blank when there is no source. Use the
{{SOURCE_LINE}}variable: fill in when there is a source<span class="info-source">来源文字</span>, otherwise an empty string. Applies to the-l、-i、-v、-c、-wmold (-mmulti-card has no footer, does not apply).
Delivery
- Report the file path
Taste guidelines
Shared by all molds. Before executing any mold, first Read references/taste.md, as the visual quality baseline running through the entire process.
Core: anti-AI-generation traces — no Inter font, no pure black, no three-equal-column cards, no centered Hero, no AI copywriting tone, no fake data.
Execution
Select the mold according to the parameters, Read references/taste.md + the corresponding mode file, and execute according to the steps:
-l (default): long image
Read references/mode-long.md, execute according to its steps.
Template:assets/long_template.html
-i: infographic
Read references/mode-infograph.md, execute according to its steps.
Template:assets/infograph_template.html
-m: multi-card
Read references/mode-poster.md, execute according to its steps.
Template:assets/poster_template.html
-v: visual notes
Read references/mode-sketchnote.md, execute according to its steps.
Template:assets/sketchnote_template.html
-c: comic
Read references/mode-comic.md, execute according to its steps.
Template:assets/comic_template.html
-w: whiteboard
Read references/mode-whiteboard.md, execute according to its steps.
Template:assets/whiteboard_template.html
-b: large text
Read references/mode-big.md, execute according to its steps.
Template:assets/big_template.html