Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when user wants a standalone HTML diagram in flat engineering blueprint style — architecture diagrams, system flows, technical spec sheets, component maps. Generates one HTML file using Tailwind v4 (browser CDN) for layout and D3 v7 (CDN) for SVG diagrams. User-invoked only — do NOT auto-trigger. Triggers on "/html-draft", "сделай blueprint", "технический чертёж", "архитектурная схема", "инженерная схема", "blueprint diagram", "engineering blueprint", "technical spec sheet", "architecture di
.claude/skills/serejaris-html-draft/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 81% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 48% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 41% | 0% |
Generate one HTML page that renders a technical diagram in a strict flat-blueprint aesthetic — the look of a printed engineering specification sheet, not a marketing landing.
Stack: Tailwind v4 via @tailwindcss/browser CDN for layout + utilities, D3 v7 via jsDelivr CDN for SVG-based diagrams (nodes, connectors, layouts, animations).
Use this when the user wants an architecture diagram, system flow, technical spec sheet, or component map as a standalone HTML artifact (suitable for slides, reports, exports).
Don't use this for:
Precise. Objective. High data-ink ratio (Tufte). Every pixel earns its place; nothing decorative. The stack is modern (Tailwind + D3) but the output looks like a printed engineering doc.
@theme)css@theme { --color-c-bg: #f8fafc; /* page background — slate-50 */ --color-c-canvas: #ffffff; /* diagram canvas */ --color-c-border: #cbd5e1; /* slate-300 */ --color-c-text-main: #0f172a; /* slate-900 */ --color-c-text-sub: #64748b; /* slate-500 */ --color-c-accent: #b91c1c; /* red-700 — semantic only */ --font-ui: system-ui, -apple-system, 'Segoe UI', sans-serif; --font-mono: 'SF Mono', Monaco, Consolas, monospace; }
Tokens become Tailwind utilities automatically: bg-c-canvas, border-c-border, text-c-text-sub, font-mono.
font-ui)font-mono.diagram-canvas — bordered box with generous padding (p-8 or more)grid / flex utilities; no eyeballingborder-t / border-l for orthogonal CSS connectors<svg>https://cdn.jsdelivr.net/npm/@tailwindcss/browser@4https://cdn.jsdelivr.net/npm/d3@7<!DOCTYPE html> through </html><style type="text/tailwindcss"> @theme block — no scattered custom CSShtml<!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8" /> <meta name="viewport" content="width=device-width, initial-scale=1.0" /> <title>[Diagram Title]</title> <script src="https://cdn.jsdelivr.net/npm/@tailwindcss/browser@4"></script> <style type="text/tailwindcss"> @theme { --color-c-bg: #f8fafc; --color-c-canvas: #ffffff; --color-c-border: #cbd5e1; --color-c-border-strong: #94a3b8; --color-c-text-main: #0f172a; --color-c-text-sub: #64748b; --color-c-accent: #b91c1c; --font-ui: system-ui, -apple-system, 'Segoe UI', sans-serif; --font-mono: 'SF Mono', Monaco, Consolas, monospace; } body { font-family: var(--font-ui); } .mono { font-family: var(--font-mono); } </style> </head> <body class="bg-c-bg text-c-text-main p-10"> <div class="max-w-[1200px] mx-auto bg-c-canvas border-2 border-c-border-strong p-8"> <header class="border-b border-c-border pb-4 mb-6 flex items-end justify-between"> <div> <h1 class="text-2xl font-semibold">[Title]</h1> <p class="mono text-[11px] uppercase tracking-widest text-c-text-sub mt-1"> [SUBTITLE] </p> </div> <div class="mono text-[11px] text-c-text-sub text-right"> DOC-[ID]<br/>REV A </div> </header> <!-- Diagram content: Tailwind grid for spec sheets, D3 SVG for flows --> <section class="grid grid-cols-2 border border-c-border"> <!-- spec cells, see snippets below --> </section> <!-- D3 mount point for SVG diagrams --> <svg id="d3-diagram" class="w-full border border-c-border mt-6" height="400"></svg> </div> <script src="https://cdn.jsdelivr.net/npm/d3@7"></script> <script> // D3 diagram rendering — see "D3 patterns" section below </script> </body> </html>
html<div class="bg-c-canvas border border-c-border p-3"> <div class="text-[10px] uppercase tracking-wide text-c-text-sub">label</div> <div class="mono text-sm">value</div> </div>
html<span class="inline-block mono text-[10px] uppercase px-1.5 py-0.5 border border-c-text-main"> ACTIVE </span> <span class="inline-block mono text-[10px] uppercase px-1.5 py-0.5 bg-c-text-main text-c-canvas"> SCHEDULED </span> <span class="inline-block mono text-[10px] uppercase px-1.5 py-0.5 bg-c-accent text-c-canvas"> OVERDUE </span>
html<div class="border-t border-c-border"></div> <div class="border-t border-dashed border-c-border"></div>
html<div class="p-4 border-r border-b border-c-border last:border-r-0"> <div class="text-[10px] uppercase tracking-wide text-c-text-sub mb-1">label</div> <div class="mono text-sm">value</div> </div>
Use D3 when geometry is non-orthogonal, computed, or large enough that hand-placing nodes is unmaintainable.
javascriptconst svg = d3.select('#d3-diagram'); const nodes = [ { id: 'api', x: 100, y: 100, label: 'API' }, { id: 'worker', x: 400, y: 100, label: 'Worker' }, { id: 'db', x: 250, y: 280, label: 'DB' }, ]; const links = [ { source: 'api', target: 'worker', style: 'solid' }, { source: 'api', target: 'db', style: 'solid' }, { source: 'worker', target: 'db', style: 'dashed' }, ]; const byId = Object.fromEntries(nodes.map(n => [n.id, n])); // arrow marker svg.append('defs').append('marker') .attr('id', 'arrow').attr('viewBox', '0 -5 10 10') .attr('refX', 8).attr('refY', 0).attr('markerWidth', 6).attr('markerHeight', 6) .attr('orient', 'auto') .append('path').attr('d', 'M0,-4L8,0L0,4').attr('fill', '#0f172a'); // links svg.selectAll('line').data(links).enter().append('line') .attr('x1', d => byId[d.source].x).attr('y1', d => byId[d.source].y) .attr('x2', d => byId[d.target].x).attr('y2', d => byId[d.target].y) .attr('stroke', '#0f172a').attr('stroke-width', 1) .attr('stroke-dasharray', d => d.style === 'dashed' ? '4 3' : null) .attr('marker-end', 'url(#arrow)'); // nodes const g = svg.selectAll('g.node').data(nodes).enter().append('g') .attr('transform', d => `translate(${d.x - 50}, ${d.y - 18})`); g.append('rect').attr('width', 100).attr('height', 36) .attr('fill', '#fff').attr('stroke', '#0f172a').attr('stroke-width', 1); g.append('text').attr('x', 50).attr('y', 22) .attr('text-anchor', 'middle').attr('font-size', 12) .attr('font-family', 'system-ui').text(d => d.label);
d3.hierarchy() + d3.tree() for parent/child trees (component maps, org charts). Render with the same flat node style; never use the default rounded D3 examples.
d3-dag (optional) or manual topological layout. For < 15 nodes, hand-place coordinates — it's faster and tighter than a layout algorithm.
d3-sankey plugin (https://cdn.jsdelivr.net/npm/d3-sankey@0.12) when volumes matter. Keep ribbons grayscale; one accent only for the watched flow.
@tailwindcss/browser@4, d3@7)mermaid-diagrams skill if the user wants MermaidMethodology adapted from QoderWork's drafter-diagram skill (flat-engineering-blueprint visual system), restacked on Tailwind v4 + D3 v7.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 50,126 | 54,997 | +10% | 1 | 1 | 0% | 8,256 | 11,603 | +41% | 0 | 0 | — |
case-02 | fail→pass | 36,853 | 43,026 | +17% | 1 | 1 | 0% | 8,262 | 11,609 | +41% | 0 | 0 | — |
case-03 | fail→fail | 42,564 | 40,392 | -5% | 1 | 1 | 0% | 8,242 | 11,589 | +41% | 0 | 0 | — |
case-04 | fail→fail | 40,750 | 51,670 | +27% | 1 | 1 | 0% | 8,243 | 11,590 | +41% | 0 | 0 | — |
case-05 | fail→fail | 48,274 | 47,745 | -1% | 1 | 1 | 0% | 8,221 | 11,568 | +41% | 0 | 0 | — |
case-06 | fail→fail | 39,696 | 37,465 | -6% | 1 | 1 | 0% | 8,219 | 11,566 | +41% | 0 | 0 | — |
case-07 | fail→pass | 38,811 | 38,585 | -1% | 1 | 1 | 0% | 8,233 | 11,580 | +41% | 0 | 0 | — |
case-08 | pass→pass | 39,274 | 42,439 | +8% | 1 | 1 | 0% | 8,229 | 11,576 | +41% | 0 | 0 | — |
case-09 | fail→fail | 47,749 | 40,606 | -15% | 1 | 1 | 0% | 8,223 | 11,570 | +41% | 0 | 0 | — |
case-10 | fail→pass | 27,483 | 50,173 | +83% | 1 | 1 | 0% | 6,389 | 11,566 | +81% | 0 | 0 | — |
case-11 | pass→pass | 48,620 | 39,907 | -18% | 1 | 1 | 0% | 8,229 | 11,576 | +41% | 0 | 0 | — |
case-12 | fail→fail | 34,852 | 37,980 | +9% | 1 | 1 | 0% | 8,217 | 11,564 | +41% | 0 | 0 | — |
case-13 | fail→fail | 27,902 | 48,797 | +75% | 1 | 1 | 0% | 6,086 | 11,565 | +90% | 0 | 0 | — |
case-14 | fail→pass | 34,956 | 50,515 | +45% | 1 | 1 | 0% | 7,807 | 11,560 | +48% | 0 | 0 | — |
case-15 | fail→pass | 38,665 | 37,846 | -2% | 1 | 1 | 0% | 8,213 | 11,560 | +41% | 0 | 0 | — |
case-16 | pass→fail | 50,528 | 37,912 | -25% | 1 | 1 | 0% | 8,215 | 11,562 | +41% | 0 | 0 | — |
case-17 | fail→pass | 36,920 | 38,706 | +5% | 1 | 1 | 0% | 8,214 | 11,561 | +41% | 0 | 0 | — |
case-18 | fail→fail | 46,770 | 39,623 | -15% | 1 | 1 | 0% | 8,214 | 11,561 | +41% | 0 | 0 | — |
case-19 | fail→fail | 45,212 | 37,886 | -16% | 1 | 1 | 0% | 7,556 | 11,559 | +53% | 0 | 0 | — |
case-20 | pass→pass | 9,664 | 11,813 | +22% | 1 | 1 | 0% | 1,700 | 5,418 | +219% | 0 | 0 | — |
case-21 | pass→pass | 50,999 | 49,846 | -2% | 1 | 1 | 0% | 8,241 | 11,588 | +41% | 0 | 0 | — |
case-22 | fail→fail | 52,592 | 41,691 | -21% | 1 | 1 | 0% | 8,227 | 11,574 | +41% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +23 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.