Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A crisp greybox / blueprint lo-fi wireframe — neutral grey blocks on a pale page, image placeholders drawn as a rectangle with a diagonal X, text shown as solid "lorem bars" of varying widths, sharp 1.5–2px borders, and a single monospace redline accent for annotations. Reads like a structural blueprint before any visual design is applied. Use when the brief asks for "greybox", "blueprint wireframe", "lo-fi dashboard", "low fidelity", "线框图", or "灰盒原型".
.claude/skills/nexu-io-wireframe-greybox/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 233% | 0% |
| case-06 | ✓→✗ | ▼ Worse | 9% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 23% | 0% |
| case-09 | ✓→✗ | ▼ Worse | 9% | 0% |
Produce a single crisp greybox wireframe page. The whole point is structure, not skin — communicate hierarchy and layout with neutral grey blocks, never with real color, copy, or imagery. Everything is a placeholder: text is grey "lorem bars", images are rectangles with a diagonal cross, numbers are big solid blocks. The only color on the page is the single redline accent reserved for monospace annotations.
this is deliberately low fidelity. Use a clean system sans (Inter / system-ui) for shape and a mono (IBM Plex Mono / JetBrains Mono) for the redline annotations only. No hand-drawn fonts, no brand color.
#f7f7f8),blocks are medium grey (#e3e3e6), borders are crisp and defined (#c9c9cf for soft edges, #1c1b1a for emphasis), and exactly one accent (a blue or coral redline) carries every monospace annotation. Keep contrast high enough to read as a small thumbnail — never near-white-on-white.
data-od-id:avatar circle on the right.
active with a filled/inverted block.
solid "number" bar.
hatched/grey bars or a schematic line. No real data.
with one column of image-placeholders drawn as a box with a diagonal X.
canvas (e.g. 01 · 12-col grid, max-width 1280).
<!doctype html>through </html>, CSS inline, no external JS or images. SVG and CSS draw every placeholder. A max-width container plus a media query collapses columns under ~1000px.
blocks, X-crossed image boxes, lorem bars, and exactly one accent color on the monospace redlines. If any greybox is missing, the page looks colored, or it disappears as a blank thumbnail, fix it.
Emit between <artifact> tags:
<artifact identifier="wireframe-slug" type="text/html" title="Wireframe — Title">
<!doctype html>
<html>...</html>
</artifact>One sentence before the artifact, nothing after.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→fail | 14,253 | 52,325 | +267% | 1 | 1 | 0% | 2,684 | 8,946 | +233% | 0 | 0 | — |
case-06 | pass→fail | 48,091 | 51,160 | +6% | 1 | 1 | 0% | 8,249 | 8,963 | +9% | 0 | 0 | — |
case-07 | fail→fail | 35,179 | 38,394 | +9% | 1 | 1 | 0% | 8,262 | 8,976 | +9% | 0 | 0 | — |
case-01 | fail→pass | 36,094 | 63,054 | +75% | 1 | 1 | 0% | 8,265 | 8,979 | +9% | 0 | 0 | — |
case-02 | fail→fail | 38,527 | 39,028 | +1% | 1 | 1 | 0% | 8,261 | 8,975 | +9% | 0 | 0 | — |
case-03 | fail→fail | 66,920 | 51,510 | -23% | 1 | 1 | 0% | 8,261 | 8,975 | +9% | 0 | 0 | — |
case-04 | pass→fail | 31,419 | 76,319 | +143% | 1 | 1 | 0% | 7,299 | 8,981 | +23% | 0 | 0 | — |
case-08 | pass→pass | 30,738 | 59,538 | +94% | 1 | 1 | 0% | 5,174 | 8,978 | +74% | 0 | 0 | — |
case-09 | pass→fail | 37,013 | 42,870 | +16% | 1 | 1 | 0% | 8,263 | 8,977 | +9% | 0 | 0 | — |
case-10 | pass→pass | 49,817 | 40,461 | -19% | 1 | 1 | 0% | 8,271 | 8,985 | +9% | 0 | 0 | — |
case-11 | pass→fail | 49,076 | 47,234 | -4% | 1 | 1 | 0% | 8,257 | 8,971 | +9% | 0 | 0 | — |
case-12 | pass→fail | 28,204 | 44,220 | +57% | 1 | 1 | 0% | 4,959 | 8,969 | +81% | 0 | 0 | — |
case-13 | pass→pass | 22,280 | 51,583 | +132% | 1 | 1 | 0% | 5,065 | 8,966 | +77% | 0 | 0 | — |
case-14 | pass→pass | 11,512 | 38,468 | +234% | 1 | 1 | 0% | 2,733 | 7,026 | +157% | 0 | 0 | — |
case-15 | pass→fail | 30,855 | 36,866 | +19% | 1 | 1 | 0% | 7,159 | 8,964 | +25% | 0 | 0 | — |
case-16 | pass→fail | 21,581 | 49,056 | +127% | 1 | 1 | 0% | 5,094 | 8,971 | +76% | 0 | 0 | — |
case-17 | pass→pass | 37,149 | 68,315 | +84% | 1 | 1 | 0% | 8,265 | 8,979 | +9% | 0 | 0 | — |
case-18 | pass→pass | 39,560 | 38,877 | -2% | 1 | 1 | 0% | 8,251 | 8,965 | +9% | 0 | 0 | — |
case-19 | pass→pass | 44,890 | 49,342 | +10% | 1 | 1 | 0% | 7,710 | 8,977 | +16% | 0 | 0 | — |
case-20 | fail→fail | 74,940 | 68,796 | -8% | 1 | 1 | 0% | 8,251 | 8,965 | +9% | 0 | 0 | — |
case-21 | pass→fail | 40,936 | 41,477 | +1% | 1 | 1 | 0% | 7,140 | 8,979 | +26% | 0 | 0 | — |
case-22 | fail→fail | 37,922 | 56,518 | +49% | 1 | 1 | 0% | 8,256 | 8,970 | +9% | 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 -67 percentage points is the difference between those two pass rates over the 22 comparable cases. 10 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.