Skills That Answer in HTML — The Highest-Scoring Pattern We Hadn't Named
Ask an agent a question and you get prose. Ask an agent with the right skill and you get a page you can show someone. The second kind is quietly dominating our top scores — and the reason is more structural than aesthetic.
Sort our directory by score and look past the usual suspects — the Office trio, the API references — and a pattern shows up that we hadn't named until we went looking: skills whose entire job is converting an answer into a visual artifact. Not "drawing tools," not "image generation" — they take the answer the agent was going to give anyway and change its output format.
Five of them score between 9.4 and 9.6:
The pattern, in five skills
answer-me-with-html — 9.6 · Developer Tools
Turns an answer into a one-page visual HTML explainer. The skill doesn't add knowledge — it adds a rendering contract: what a good explainer page contains, how it's structured, what makes it self-contained. Nearly 3,000 stars on its source repo, and the score suggests the stars are earned.
anatomy — 9.5 · Developer Tools
Builds crafted, highly detailed, interactive isometric SVG figures. The output is an SVG file — but the real payload is the craft standard: what "highly detailed" means, how an isometric figure is composed, what passes and what's lazy.
baoyu-infographic — 9.3 · Design
Professional infographics with 21 layout types. The layout library is what makes it a skill rather than a prompt — the agent doesn't invent an infographic structure, it selects from tested ones.
anything2explainer — 9.4 · Video
Give it a topic, get a black-background motion-graphics explainer video with narration, subtitles, chapters, and a progress bar. The most ambitious of the five — it converts an explanation into a produced artifact, not just a styled one.
baoyu-diagram — 9.4 · Design
Professional dark-themed SVG diagrams of any type. Same shape as anatomy, different domain.
Why this scores so well
We score six dimensions per skill. Two of them — workflow design and content quality — are where these five separate from average, and both for the same underlying reason:
A rendering contract is checkable. "Write a good explanation" gives an agent nothing to verify against. "The explainer must fit one page, have a visual anchor in the first viewport, and not exceed N sections" gives it a checklist. Skills that encode checkable contracts produce output the agent can self-review before showing you — and in our data, that property alone moves scores by half a point or more.
The output format forces organization. Prose answers hide lazy thinking; a visual artifact exposes it. You cannot make an infographic without deciding what the three most important things are. The format does the prioritization work that a plain answer skips.
The design lesson
If you're writing a skill, this pattern is stealable: pick the answer your users already get, and write the contract for a better rendering of it. The three questions these skills all answer:
- What are the structural elements? (sections, layout types, figure components)
- What makes each element pass or fail? (the craft standard)
- What's the self-check before delivery? (fits one page, renders standalone, has an anchor)
Answer those in a SKILL.md and you're most of the way to a 9.
Where to browse
The design and video categories hold the rest of this pattern — infographics, motion graphics, 3D explainers, diagramming. And if you want the underlying method for reading any of these files quickly, our 60-second SKILL.md reading routine covers what to check first.