Claude Skill

whatis

What Is This — explain anything as a picture-first HTML page. Use when the user types /whatis <topic>, or asks what something is / how it works and would be better served by big diagrams than by paragraphs.

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Download suyuan2022-suyuan-skill-whatis-5a6848f.zip · 1 KB

Install

skills CLI npx skills add https://github.com/suyuan2022/suyuan-skill/tree/main/whatis
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install suyuan2022-suyuan-skill@llmmart
Git git clone https://github.com/suyuan2022/suyuan-skill.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole suyuan2022/suyuan-skill collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

whatis

Explain like I'm someone who knows nothing about this topic, using an HTML page with big pictures and few words.

No Artifact tool in this environment: write a self-contained HTML file (inline SVG/CSS, no external deps) to a temp directory (e.g. $TMPDIR or /tmp), then open it.

If the current context is not enough to explain the topic, gather enough context yourself before drawing. Every node in a diagram must be backed by evidence; never fabricate.

Draw real diagrams in SVG, freely picking and combining whatever fits: flowchart, sequence diagram, state diagram, architecture/block diagram, C4, dependency graph, call graph, ER diagram, class diagram, DFD, swimlane, deployment diagram, timeline, Gantt, git graph, before/after comparison, bar/line chart, treemap, heatmap... If none fits, design your own — expression comes before templates. Tables are for enumerations only and don't count as diagrams.

A diagram exists to make structure visible: who talks to whom, what causes what, what changes when. Draw each one in the shape the reader's eye already knows — boxes for parts, arrows for actions, lanes for actors, time flowing one way. Make each diagram read at two distances: at a glance only the backbone shows — a name per box, a verb per arrow, one path the eye can follow; the detail lives in a quieter layer of smaller, dimmer annotations that never competes with the backbone. Color like a print magazine: warm paper background, ink for text and lines, one muted accent doing all the pointing; status reads from marks and line weight, tinted from the same palette. One diagram, one point: when the content outgrows the shape, split it or summarize the tail — shrinking the type is never the answer. The test: someone who reads only the diagrams and captions still gets the whole story.

Give the page a wide content column with visible margins on both sides: diagrams span the column (each SVG gets a viewBox and width:100% so it scales with the window), prose sits in a narrower measure within it.

Match the user's language in the explanation. Keep technical terms, API names, CLI commands, and exact error strings verbatim in English.

Files (suyuan-skill)
  • SKILL.md 2.4 KB
    ---
    name: whatis
    description: "What Is This — explain anything as a picture-first HTML page. Use when the user types /whatis <topic>, or asks what something is / how it works and would be better served by big diagrams than by paragraphs."
    ---
    
    # whatis
    
    Explain like I'm someone who knows nothing about this topic, using an HTML page with big pictures and few words.
    
    No Artifact tool in this environment: write a self-contained HTML file (inline SVG/CSS, no external deps) to a temp directory (e.g. `$TMPDIR` or `/tmp`), then `open` it.
    
    If the current context is not enough to explain the topic, gather enough context yourself before drawing. Every node in a diagram must be backed by evidence; never fabricate.
    
    Draw real diagrams in SVG, freely picking and combining whatever fits: flowchart, sequence diagram, state diagram, architecture/block diagram, C4, dependency graph, call graph, ER diagram, class diagram, DFD, swimlane, deployment diagram, timeline, Gantt, git graph, before/after comparison, bar/line chart, treemap, heatmap... If none fits, design your own — expression comes before templates. Tables are for enumerations only and don't count as diagrams.
    
    A diagram exists to make structure visible: who talks to whom, what causes what, what changes when. Draw each one in the shape the reader's eye already knows — boxes for parts, arrows for actions, lanes for actors, time flowing one way. Make each diagram read at two distances: at a glance only the backbone shows — a name per box, a verb per arrow, one path the eye can follow; the detail lives in a quieter layer of smaller, dimmer annotations that never competes with the backbone. Color like a print magazine: warm paper background, ink for text and lines, one muted accent doing all the pointing; status reads from marks and line weight, tinted from the same palette. One diagram, one point: when the content outgrows the shape, split it or summarize the tail — shrinking the type is never the answer. The test: someone who reads only the diagrams and captions still gets the whole story.
    
    Give the page a wide content column with visible margins on both sides: diagrams span the column (each SVG gets a viewBox and `width:100%` so it scales with the window), prose sits in a narrower measure within it.
    
    Match the user's language in the explanation. Keep technical terms, API names, CLI commands, and exact error strings verbatim in English.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.