Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →网页转 Markdown - 抓取任意网页正文并转成干净的 Markdown，便于阅读/收藏/二次创作
.claude/skills/dongsheng123132-web-to-markdown/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -83% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -18% | 0% |
把用户给的网页链接抓下来，提取正文，转成干净的 Markdown，方便保存、总结或二次创作。
用 Bash 工具。首选 jina.ai 的免费 reader（无需依赖，最省事）：
bash# 最简：jina reader 直接返回干净 Markdown（在链接前加 https://r.jina.ai/） curl -s "https://r.jina.ai/https://example.com/article" -o article.md echo "已保存 -> article.md"; head -40 article.md
离线或 jina 不可用时，用 Python 本地转换：
bashpython -c "import markdownify,requests" 2>/dev/null || pip install -q markdownify requests beautifulsoup4 python - <<'PY' import requests, re from bs4 import BeautifulSoup from markdownify import markdownify as md url = "https://example.com/article" html = requests.get(url, timeout=15, headers={"User-Agent":"Mozilla/5.0"}).text soup = BeautifulSoup(html, "html.parser") for t in soup(["script","style","nav","footer","aside"]): t.decompose() body = soup.find("article") or soup.find("main") or soup.body out = md(str(body), heading_style="ATX") out = re.sub(r"\n{3,}", "\n\n", out).strip() open("article.md","w",encoding="utf-8").write(out) print("已保存 -> article.md，", len(out), "字") PY
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→fail | 16,107 | 9,076 | -44% | 1 | 1 | 0% | 3,018 | 954 | -68% | 0 | 0 | — |
case-01 | fail→fail | 11,423 | 38,220 | +235% | 1 | 1 | 0% | 2,150 | 1,146 | -47% | 0 | 0 | — |
case-02 | fail→fail | 25,699 | 9,534 | -63% | 1 | 1 | 0% | 2,686 | 1,183 | -56% | 0 | 0 | — |
case-03 | fail→fail | 18,722 | 7,531 | -60% | 1 | 1 | 0% | 3,410 | 908 | -73% | 0 | 0 | — |
case-04 | pass→pass | 9,451 | 4,428 | -53% | 1 | 1 | 0% | 1,437 | 1,184 | -18% | 0 | 0 | — |
case-05 | pass→pass | 14,784 | 13,184 | -11% | 1 | 1 | 0% | 2,204 | 2,677 | +21% | 0 | 0 | — |
case-06 | pass→pass | 5,711 | 5,789 | +1% | 1 | 1 | 0% | 1,001 | 1,280 | +28% | 0 | 0 | — |
case-08 | pass→pass | 7,786 | 5,493 | -29% | 1 | 1 | 0% | 1,123 | 1,303 | +16% | 0 | 0 | — |
case-09 | pass→pass | 9,682 | 4,730 | -51% | 1 | 1 | 0% | 1,807 | 1,196 | -34% | 0 | 0 | — |
case-10 | fail→pass | 4,898 | 2,025 | -59% | 1 | 1 | 0% | 646 | 798 | +24% | 0 | 0 | — |
case-11 | pass→pass | 7,733 | 4,818 | -38% | 1 | 1 | 0% | 1,066 | 1,086 | +2% | 0 | 0 | — |
case-12 | fail→pass | 7,159 | 5,510 | -23% | 1 | 1 | 0% | 1,230 | 1,277 | +4% | 0 | 0 | — |
case-13 | pass→pass | 8,960 | 4,176 | -53% | 1 | 1 | 0% | 713 | 1,000 | +40% | 0 | 0 | — |
case-14 | pass→pass | 6,173 | 6,262 | +1% | 1 | 1 | 0% | 1,053 | 1,373 | +30% | 0 | 0 | — |
case-15 | pass→pass | 4,980 | 4,101 | -18% | 1 | 1 | 0% | 853 | 1,175 | +38% | 0 | 0 | — |
case-16 | pass→pass | 8,658 | 3,609 | -58% | 1 | 1 | 0% | 1,396 | 1,006 | -28% | 0 | 0 | — |
case-17 | pass→pass | 6,949 | 3,044 | -56% | 1 | 1 | 0% | 1,107 | 992 | -10% | 0 | 0 | — |
case-18 | fail→pass | 8,489 | 2,059 | -76% | 1 | 1 | 0% | 1,379 | 810 | -41% | 0 | 0 | — |
case-19 | pass→pass | 8,273 | 2,864 | -65% | 1 | 1 | 0% | 1,060 | 1,020 | -4% | 0 | 0 | — |
case-20 | pass→pass | 11,852 | 10,511 | -11% | 1 | 1 | 0% | 2,287 | 2,152 | -6% | 0 | 0 | — |
case-21 | pass→fail | 18,770 | 10,871 | -42% | 1 | 1 | 0% | 3,802 | 665 | -83% | 0 | 0 | — |
case-22 | pass→pass | 16,259 | 15,317 | -6% | 1 | 1 | 0% | 2,922 | 3,362 | +15% | 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 +9 percentage points is the difference between those two pass rates over the 17 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.