HTML output skill for Claude agents — 4 levels (Static Doc / Visual Artifact / Two-Way Interactive / Throwaway Tool). Based on Thariq Shihipar's framework.
개요
A Claude skill that teaches your agent to produce — across four levels of capability. Based on Thariq Shihipar's article Using Claude Code: The Unreasonable Effectiveness of HTML. Go read the source — this skill is one community implementation of his ideas. When you ask your agent for a doc, plan, report, review, design, dashboard, or editor — it builds an HTML artifact at the right level instead of dumping markdown. Each level adds capability and effort. Most agent output stops at Level 1 — the unlock is recognising when you actually need 3 or 4. Then in Claude Code, the skill triggers on: - /html-it - "html-it" / "html this" / "render as html" - "build me an html for X" / "make an html artifact" Read this folder and produce an HTML implementation plan — TL;DR at top, key code snippets inline, sticky TOC on the left. Easy to digest on mobile. Read the rate limiter code.
README
html-it
A Claude skill that teaches your agent to produce HTML output instead of markdown — across four levels of capability.
Based on Thariq Shihipar’s article Using Claude Code: The Unreasonable Effectiveness of HTML. Go read the source — this skill is one community implementation of his ideas.
Live examples gallery (Thariq’s): thariqs.github.io/html-effectiveness
What it does
When you ask your agent for a doc, plan, report, review, design, dashboard, or editor — it builds an HTML artifact at the right level instead of dumping markdown.
The four levels:
| Level | What | Examples |
|---|---|---|
| 1 — Static Doc | HTML as a prettier markdown | Specs, reports, explainers, plans |
| 2 — Visual Artifact | Adds SVG, tables, layouts | Design systems, decks, dashboards |
| 3 — Two-Way Interactive | Sliders/toggles + export button | Tuning, A/B picks, prompt iteration |
| 4 — Throwaway Tool | Purpose-built mini-app | Triage boards, config editors |
Each level adds capability and effort. Most agent output stops at Level 1 — the unlock is recognising when you actually need 3 or 4.
Install
In your Claude Code project:
git clone https://github.com/robonuggets/html-it ~/.claude/skills/html-it
Or for a project-specific install:
cd your-project
mkdir -p .claude/skills
git clone https://github.com/robonuggets/html-it .claude/skills/html-it
Then in Claude Code, the skill triggers on:
/html-it- “html-it” / “html this” / “render as html”
- “build me an html for X” / “make an html artifact”
Quick prompts (one per level)
Level 1 (static doc):
Read this folder and produce an HTML implementation plan — TL;DR at top, key code snippets inline, sticky TOC on the left. Easy to digest on mobile.
Level 2 (visual artifact):
Read the rate limiter code. Produce a single HTML explainer page: token-bucket SVG flow at the top, 3-4 annotated code snippets, gotchas section at the bottom.
Level 3 (two-way interactive):
I’m tuning this system prompt. Make a side-by-side HTML editor — editable prompt on the left with variable slots, three sample inputs on the right that re-render live. Add a character counter and a Copy as prompt button.
Level 4 (throwaway tool):
I need to reprioritise these 30 tickets. Make an HTML file with each ticket as a draggable card across Now / Next / Later / Cut columns. Pre-sort by your best guess. Add a Copy as markdown button that exports the final ordering with a one-line rationale per bucket.
What’s in the box
SKILL.md— the skill itself. Drop into.claude/skills/html-it/and Claude picks it up.README.md— this file.
The skill encodes:
- When to choose HTML over markdown
- The four levels and which patterns belong to each
- A locked design system (ivory/slate/clay palette, serif+sans+mono stack) so output doesn’t look AI-slop
- The mandatory “export back to prompt” pattern for Level 3+
- Anti-patterns to avoid
Related skills
tweak— Inject a live tweak panel into any HTML (5 or 10 sliders), bake selected values back into source CSS. Pairs naturally with Level 3/4 outputs fromhtml-it.
Credit
All credit to Thariq Shihipar of the Anthropic Claude Code team for the framework and the use-case taxonomy. The four-level escalation is the maintainer’s reading of his article — defer to his original write-up when in doubt.
Article: Using Claude Code — The Unreasonable Effectiveness of HTML Examples: thariqs.github.io/html-effectiveness
License
MIT. Use it, fork it, improve it.
Built by RoboNuggets.
추천 도구
다른 키워드를 입력하거나 필터를 제거해 보세요.
설치
npx skillfish add robonuggets/html-it