Clean, single-file HTML for your coding agents.
A lightweight skill that teaches agents to output clean, zero-dependency HTML.
No npm, no CDNs, 100% offline. Built-in shadcn-inspired neutrals so reports and dashboards don't look like AI slop.
Basic Components & Vector Icons
Test the live atomic components and copy-paste snippets directly into your projects.
Expandable <details> Accordion ▾
| Trace ID | Service Cluster | Latency | Memory | Status |
|---|---|---|---|---|
tr-0812-alpha |
Auth-Gateway-Core | 14ms | 128 MB | Healthy |
tr-0812-beta |
Inference-Worker-Pool | 420ms | 4.2 GB | Optimized |
tr-0812-gamma |
Async-Redis-Pipeline | 1.2s | 890 MB | Spike |
tr-0812-delta |
Telemetry-Ingestion | 88ms | 340 MB | Streaming |
stroke="currentColor" that automatically inherit font color. Zero web fonts, 100% offline.7 Common Layout Templates
Pick a layout that fits your output instead of letting models invent messy ad-hoc structures.
Single-Column Document & Executive Report
Centered 860px reading container optimized for executive reviews, audit memos, architecture RFCs, and print-to-PDF workflows. Features metadata headers, score cards, callouts, and native details accordions.
API Usage & Cost Dashboard & Real-Time Data Grid
Wide-screen fluid layout with a 4-column KPI stat grid, pure-SVG area and bar charts, dual-filter toolbar (search + status dropdown), and a native modal dialog for create/edit operations.
Master-Detail Split Workbench (100vh)
App-like layout occupying full viewport height without page scroll. Left sidebar filters items, right detail pane updates dynamically with 3-column property grids, verdict actions, and syntax-styled code blocks.
DeepSeek vs GPT-4o Comparison & A/B Eval Matrix
Two-column split view (Baseline vs. Challenger) with verdict callout, parameter specs, sample payload outputs, and a quantitative delta matrix table. Perfect for model evals and prompt revisions.
Vertical Service Recovery Timeline & Incident Postmortem
High-density single-track vertical timeline with semantic status nodes (Error, Warning, Success, Info), exact timestamps, operator tags, expandable diagnosis logs, and one-click "Copy as Markdown" export.
Triage & Agile Kanban Board (Native Drag & Drop)
4-column task categorization board (Backlog, Progress, Blocked, Done). Zero external dependencies, pure HTML5 drag-and-drop. Features a "Copy Triage Decisions to Agent" button to close the interactive human-in-the-loop workflow.
Diff-Based Code Review & Per-File Verdicts
Paste raw git diff output and it renders as a line-numbered unified diff. Findings are pinned to exact lines with severity and suggested changes, each file gets a verdict, and one click copies the whole review back to the agent.
Why not a CDN, or just let the AI freestyle CSS?
Three common ways agents ship HTML, and how they hold up.
<script src="cdn.tailwindcss.com">.- ✕ Blank page offline or on an intranet
- ✕ Visible lag and layout flash on open
- ✕ Easily blocked by Content Security Policy
- ✕ Old files break once the CDN link dies
- ✕ Harsh borders, chaotic spacing, no taste
- ✕ Cramped tables, unreadable on mobile
- ✕ No dark mode, glaring at night
- ✕ Random class names every run
- ✓ Fully offline single file, opens instantly
- ✓ Zero dependencies, no build step
- ✓ Clean shadcn-style neutral palette
- ✓ Smooth built-in light / dark switching
Tested Agent Prompts
A few prompts you can give your coding agent to get clean visual HTML instead of a wall of markdown.
# 1. Install via skills.sh ecosystem (supports Claude Code, Pi, Cursor, Codex, etc.)
npx skills add QingYunA/agent-html -g
# 2. Or use the standalone CLI to extract templates anywhere
npx agent-html template dashboard > dashboard.html
npx agent-html check dashboard.html