Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Open Design procurement & security leave-behind: the one-pager-plus a buying committee can forward and approve internally. Built as a decision-grade B2B sales deck for buying committee, security, procurement.
.claude/skills/nexu-io-huashu-takram-soft-tech/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 71% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 34% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 119% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 48% | 0% |
Produce a single-file, soft nature-toned tech deck. You are a research studio designer working in HTML: every chart is composed like a museum plate, every color is borrowed from moss, paper and dusk. The visual system, canvas contract, and navigation runtime are locked by example.html. Start from example.html, replace content only — do not rewrite the design or the script. Do not introduce any color or font outside this spec.
Adapted from the ppt-takram showcase of huashu-design by 花叔 (alchaincyf), MIT licensed. The upstream file is a single benchmark page; this plugin extends it into a full multi-section deck while preserving its DNA: data visualization treated as an art piece.
<div id="stage"> fixed at 1920 × 1080 px, centered withposition: fixed; top: 50%; left: 50%; a fit() function applies translate(-50%, -50%) scale(min(innerWidth/1920, innerHeight/1080)) on load and resize. All inner layout in px — the scaler owns responsiveness.
<section class="slide"> inside #stage with adata-screen-label="01 封面"-style label; exactly one slide carries .active. This is an all-light style — there are no dark pages; rhythm comes from density alternation, not theme swaps.
←/↑/PageUp previous,→/↓/Space/PageDown next, Home/End first/last; #/N hash routing (1-indexed) read on load + hashchange, written via history.replaceState; click left third = back, rest = forward; fixed counter pill bottom-right, key-hint bottom-left. No external JS, no build step — the file must open inside a sandboxed iframe via file://.
:root — keep the names, re-theme values only)| Token | Value | Role | |---|---|---| | --paper | #F5F0EB | warm rice-paper page background | | --ink | #2D3436 | heading ink | | --body | #3A3A3A | body ink | | --green | #6B8F71 | sage green — the primary accent | | --green-2 | #7D9B72 | lighter sage (delta pills, fills) | | --sage | #A8B5A0 | pale sage (labels, washes, flow lines) | | --gold | #D4A574 | the only warm accent — 2nd data series | | --gray-warm | #C8C2B8 | warm gray — 3rd data series, captions | | --gray-line | #D4CFC6 | delicate axis / rule lines | | --gray-dash | #DDD9D2 | dashed grid circles | | --gray-bg | #E8E4DC | faint outermost rings | | --muted | #615C54 | secondary text (warm ink, WCAG AA on paper) | | --faint | #736D64 | tertiary captions (warm ink, WCAG AA) | | --card | rgba(255,255,255,0.6) | translucent card background | | --card-line | rgba(168,181,160,0.18) | card hairline border | | --serif | 'Noto Serif SC', 'Songti SC', serif | display headings, quotes, insight text | | --sans | 'Inter', 'PingFang SC', 'Hiragino Sans GB', sans-serif | labels, numbers, body |
Fonts come from one Google Fonts @import (Inter 300–600 + Noto Serif SC 300–700) — the only external reference allowed. No purple, no blue, no neon, no pure black, no gradients other than the two locked radial washes.
like museum plates — concentric dashed-circle grids (--gray-dash, stroke-dasharray="2,6", decreasing stroke-width outward-in), delicate 0.5px axis lines, tiny hollow endpoint circles, faint scale numbers, and a 3-series overlay: green polygon prominent (stroke 2, dot r=6 with halo ring), gold secondary (stroke 1.2, dot r=3.5), warm-gray tertiary (stroke 1 dashed, dot r=2.5). Value callouts use thin annotation lines + 14px/600 green numerals. Every plate signs off with a right-aligned 8–9px Fig. NN — Title caption in --gray-warm.
::before carries the two lockedradial-gradients (sage at 20%/50%, warm gray at 80%/30%, ≤8% alpha) — the page must breathe, never be flat white.
.soft-card): 16px radius,rgba(255,255,255,0.6) fill, --card-line hairline border, and a 32px × 2px sage tick at top-left (::before). Value typography: 38–44px weight-300 numerals with a smaller green unit span; deltas in a rounded sage pill with a tiny SVG triangle (gold pill allowed once per page for a special callout).
rgba(107,143,113,0.15); left = green dot + 11px/500/3px-tracking uppercase deck name; right = 10px/400 --faint section label. Footer: 11px faint credit line bottom-left, Fig./Sec. note bottom-right. Page numbering lives only in the runtime counter pill.
--serif weight 400–500 (never 700+ display bombast); all labels, numerals, units and captions in --sans. English subtitles are 15–18px weight-300 --muted prose under Chinese serif titles.
label on 8% green fill) and the .insight card (10px uppercase sage label + serif 17px/1.9 finding text) for key takeaways.
| Layout | Role | |---|---| | cover | left serif display title + badge, right oversized dashed-ring SVG art | | contents | serif rows + green index numbers separated by sage hairlines | | philosophy | one 60–70px serif claim + 3 soft-cards | | big-number | 240–280px weight-300 numeral + green unit + hairline stat row | | radar (the master) | 480px left panel (kicker/serif title/badge/insight/credit) + right radar plate + 3 metric cards — preserve this two-panel grid from upstream | | diagram | SVG flow/architecture plate: dashed clusters, sage dashed flow paths, gold middle node, green terminal node | | metrics-grid | 2×3 soft-card grid with delta pills | | quote | centered serif 52px quote, gold SVG quote marks, key phrase in green | | roadmap | hairline timeline, dots graded --gray-warm → --sage → --green-2 → --green (last with halo) | | closing | centered serif CTA over faint concentric rings + badge |
letter-spacing ~1px, line-height 1.35–1.65. Hero numerals 240–280px weight 300 with negative tracking.
letter-spacing: 1.5–3px, uppercase — theconnective tissue of the style. Body/captions 13–18px weight 300, line-height 1.7–1.9.
English weight-300 sans second.
plate → whitespace/quote page. At least one radar/diagram art plate per deck — it is the reason this style exists.
series, quote marks, or one special delta pill; warm gray is the 3rd series and all chrome. Never promote gold or gray to a heading color.
drop-shadow(0 4px 20px rgba(0,0,0,0.04)), no border heavier than 1px except SVG data strokes, no Chart.js/mermaid — all charts are hand-laid inline SVG.
<!-- 待用户提供 --> placeholder, never invented statistics. (The seed deck's 「青屿/QINGYU」 data is fictional demo content and must be fully replaced.)
dataset becomes the radar/diagram art plate — this style lives and dies by its chart plate; pick it before writing any page.
example.html, retitle, then replace each section's contentfollowing the layout enumeration. Keep masthead/footer chrome, washes, token names, and the script intact. Re-theme by changing token values only — and only if the user's brand genuinely demands it; the sage/gold/warm-gray triad on rice paper is the identity.
page, confirm the grammar, then batch the rest.
(cx + r·sin(θ), cy − r·cos(θ)) with r = 220 · value/100, θ ∈ {0°, 120°, 240°} around (280, 280) — printed values and polygon shape must agree proportionally.
#/N hash stay in sync; no overflow beyond 1920×1080; every chart plate has its Fig. NN caption; only the three series colors appear in any chart; no leftover demo (「青屿」/"QINGYU") text; grep for TODO.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 21,248 | 16,314 | -23% | 1 | 1 | 0% | 3,536 | 6,030 | +71% | 0 | 0 | — |
case-01 | fail→pass | 56,939 | 48,309 | -15% | 1 | 1 | 0% | 8,332 | 11,139 | +34% | 0 | 0 | — |
case-02 | fail→fail | 41,411 | 48,676 | +18% | 1 | 1 | 0% | 8,361 | 11,168 | +34% | 0 | 0 | — |
case-03 | fail→fail | 51,357 | 45,172 | -12% | 1 | 1 | 0% | 8,343 | 11,150 | +34% | 0 | 0 | — |
case-04 | fail→pass | 20,800 | 16,009 | -23% | 1 | 1 | 0% | 3,591 | 6,573 | +83% | 0 | 0 | — |
case-05 | fail→pass | 29,670 | 77,573 | +161% | 1 | 1 | 0% | 5,043 | 11,048 | +119% | 0 | 0 | — |
case-06 | pass→fail | 21,063 | 35,548 | +69% | 1 | 1 | 0% | 2,867 | 11,046 | +285% | 0 | 0 | — |
case-07 | fail→pass | 21,720 | 11,743 | -46% | 1 | 1 | 0% | 3,145 | 4,639 | +48% | 0 | 0 | — |
case-08 | fail→pass | 30,638 | 38,856 | +27% | 1 | 1 | 0% | 6,036 | 8,644 | +43% | 0 | 0 | — |
case-10 | fail→pass | 24,029 | 11,358 | -53% | 1 | 1 | 0% | 3,582 | 5,050 | +41% | 0 | 0 | — |
case-11 | fail→pass | 41,627 | 37,146 | -11% | 1 | 1 | 0% | 6,730 | 11,031 | +64% | 0 | 0 | — |
case-12 | fail→pass | 38,877 | 48,677 | +25% | 1 | 1 | 0% | 8,208 | 11,015 | +34% | 0 | 0 | — |
case-13 | fail→pass | 13,702 | 19,437 | +42% | 1 | 1 | 0% | 2,076 | 5,614 | +170% | 0 | 0 | — |
case-14 | fail→pass | 26,267 | 26,983 | +3% | 1 | 1 | 0% | 3,926 | 6,887 | +75% | 0 | 0 | — |
case-15 | fail→pass | 14,928 | 26,455 | +77% | 1 | 1 | 0% | 2,328 | 6,752 | +190% | 0 | 0 | — |
case-16 | fail→fail | 24,608 | 26,304 | +7% | 1 | 1 | 0% | 3,689 | 7,974 | +116% | 0 | 0 | — |
case-17 | fail→pass | 34,010 | 37,240 | +9% | 1 | 1 | 0% | 5,429 | 11,015 | +103% | 0 | 0 | — |
case-18 | fail→pass | 14,500 | 5,209 | -64% | 1 | 1 | 0% | 1,917 | 3,755 | +96% | 0 | 0 | — |
case-19 | fail→fail | 14,970 | 47,684 | +219% | 1 | 1 | 0% | 2,491 | 11,012 | +342% | 0 | 0 | — |
case-20 | fail→fail | 36,383 | 39,600 | +9% | 1 | 1 | 0% | 2,555 | 11,021 | +331% | 0 | 0 | — |
case-21 | fail→fail | 19,896 | 46,549 | +134% | 1 | 1 | 0% | 3,727 | 11,044 | +196% | 0 | 0 | — |
case-22 | pass→fail | 22,627 | 42,856 | +89% | 1 | 1 | 0% | 4,549 | 11,038 | +143% | 0 | 0 | — |
case-23 | pass→pass | 20,985 | 27,656 | +32% | 1 | 1 | 0% | 3,240 | 7,919 | +144% | 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. 23 cases were attempted. The headline lift of +52 percentage points is the difference between those two pass rates over the 23 comparable cases. 3 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.