Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →分析网页截图，提取设计系统（Design System）并生成结构化数据和可用的 AI Coding Prompt。适用于 UI/UX 设计师和前端工程师需要从现有网页设计中提取设计规范、配色方案、排版系统和组件风格的场景。
.claude/skills/anbeime-web-design-analyzer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 4% | 0% |
Pillow>=10.0.0
scripts/analyze_design.py --image <图片文件路径> 执行分析--input 指定 JSON 路径，--output 指定输出路径）bash # 用户：请分析这张网页截图 # 智能体调用脚本 python scripts/analyze_design.py --image ./uploads/landing-page.png
bash # 用户：分析这个网页，然后用提取的设计规范创建一个 Hero Section # 智能体： # 1. 调用脚本分析图片 # 2. 展示结果 # 3. 使用生成的 Coding Prompt 指导代码生成
bash # 用户：分析这个网页，然后导出为路演视频风格 # 智能体： # 1. 调用 analyze_design.py 分析图片，生成 design_system.json # 2. 调用 convert_to_roadshow_style.py 转换，生成 brand_style.json # 3. 提示用户可以将 brand_style.json 用于路演视频生成 # 1. 调用脚本分析图片 # 2. 展示结果 # 3. 使用生成的 Coding Prompt 指导代码生成
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 22,641 | 2,274 | -90% | 1 | 1 | 0% | 1,378 | 1,620 | +18% | 0 | 0 | — |
case-01 | fail→fail | 22,601 | 5,857 | -74% | 1 | 1 | 0% | 5,556 | 1,483 | -73% | 0 | 0 | — |
case-02 | pass→pass | 6,009 | 1,999 | -67% | 1 | 1 | 0% | 1,026 | 1,504 | +47% | 0 | 0 | — |
case-03 | fail→pass | 15,898 | 5,207 | -67% | 1 | 1 | 0% | 2,200 | 2,141 | -3% | 0 | 0 | — |
case-05 | fail→pass | 5,026 | 1,945 | -61% | 1 | 1 | 0% | 811 | 1,512 | +86% | 0 | 0 | — |
case-06 | fail→pass | 5,810 | 2,271 | -61% | 1 | 1 | 0% | 988 | 1,562 | +58% | 0 | 0 | — |
case-07 | pass→pass | 8,980 | 6,678 | -26% | 1 | 1 | 0% | 1,466 | 2,349 | +60% | 0 | 0 | — |
case-08 | pass→pass | 5,343 | 3,429 | -36% | 1 | 1 | 0% | 873 | 1,775 | +103% | 0 | 0 | — |
case-09 | fail→pass | 9,283 | 2,521 | -73% | 1 | 1 | 0% | 1,506 | 1,560 | +4% | 0 | 0 | — |
case-10 | fail→pass | 10,008 | 2,705 | -73% | 1 | 1 | 0% | 1,863 | 1,551 | -17% | 0 | 0 | — |
case-11 | pass→pass | 4,156 | 1,838 | -56% | 1 | 1 | 0% | 644 | 1,443 | +124% | 0 | 0 | — |
case-12 | fail→fail | 7,434 | 2,817 | -62% | 1 | 1 | 0% | 1,132 | 1,619 | +43% | 0 | 0 | — |
case-17 | fail→fail | 5,620 | 9,833 | +75% | 1 | 1 | 0% | 925 | 1,849 | +100% | 0 | 0 | — |
case-13 | fail→fail | 9,136 | 13,712 | +50% | 1 | 1 | 0% | 1,733 | 3,095 | +79% | 0 | 0 | — |
case-14 | fail→fail | 4,414 | 6,304 | +43% | 1 | 1 | 0% | 699 | 1,354 | +94% | 0 | 0 | — |
case-15 | fail→fail | 10,972 | 8,133 | -26% | 1 | 1 | 0% | 2,067 | 1,889 | -9% | 0 | 0 | — |
case-16 | fail→pass | 14,823 | 6,744 | -55% | 1 | 1 | 0% | 3,169 | 2,484 | -22% | 0 | 0 | — |
case-18 | pass→pass | 9,298 | 3,622 | -61% | 1 | 1 | 0% | 1,471 | 1,682 | +14% | 0 | 0 | — |
case-19 | fail→fail | 10,217 | 6,391 | -37% | 1 | 1 | 0% | 1,744 | 2,251 | +29% | 0 | 0 | — |
case-20 | fail→fail | 4,528 | 4,849 | +7% | 1 | 1 | 0% | 764 | 1,449 | +90% | 0 | 0 | — |
case-21 | fail→fail | 3,682 | 6,630 | +80% | 1 | 1 | 0% | 587 | 1,679 | +186% | 0 | 0 | — |
case-22 | pass→pass | 6,474 | 2,007 | -69% | 1 | 1 | 0% | 1,102 | 1,500 | +36% | 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, and 17 counted toward the lift figure. The other 5 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +32 percentage points is the difference between those two pass rates over the 17 comparable cases.
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.