Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →根据arXiv论文网址自动下载PDF并进行多维度分析，包括文本提取、词频分析、语音播报、播客对话生成、交互式网页、PPT、总结图和引用分析
.claude/skills/anbeime-paper-analysis-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 173% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 17% | 0% |
scripts/download_pdf.py 下载 arXiv PDF--url (arXiv 论文网址), --output (输出 PDF 文件路径)scripts/extract_text.py 提取纯文本--pdf (PDF 文件路径), --output (输出 txt 文件路径)scripts/analyze_word_frequency.py 进行词频统计--txt (txt 文件路径), --output (输出 csv 文件路径)scripts/text_to_speech.py 将文本转为语音--txt (txt 文件路径), --output (输出 wav 文件路径)scripts/dialogue_to_podcast.py 将对话脚本转换为语音--dialogue (对话脚本文件路径), --output (输出 wav 文件路径)scripts/generate_html.py 生成交互式网页--txt (txt 文件路径), --word_freq (词频 csv 文件路径), --output (输出 html 文件路径)scripts/generate_ppt.py 生成演示文稿--txt (txt 文件路径), --output (输出 pptx 文件路径)scripts/extract_references.py 提取引用链接--txt (txt 文件路径), --output (输出 csv 文件路径)bash # 下载 PDF python scripts/download_pdf.py --url "https://arxiv.org/abs/2301.00001" --output ./user-data/paper.pdf # 提取文本 python scripts/extract_text.py --pdf ./user-data/paper.pdf --output ./user-data/paper.txt # 词频分析 python scripts/analyze_word_frequency.py --txt ./user-data/paper.txt --output ./user-data/word_freq.csv # 语音合成 python scripts/text_to_speech.py --txt ./user-data/paper.txt --output ./user-data/paper.wav # 播客对话（智能体生成对话脚本后） python scripts/dialogue_to_podcast.py --dialogue ./user-data/dialogue.txt --output ./user-data/podcast.wav # 生成网页 python scripts/generate_html.py --txt ./user-data/paper.txt --word_freq ./user-data/word_freq.csv --output ./user-data/analysis.html # 生成 PPT python scripts/generate_ppt.py --txt ./user-data/paper.txt --output ./user-data/presentation.pptx # 提取引用 python scripts/extract_references.py --txt ./user-data/paper.txt --output ./user-data/references.csv
bash python scripts/analyze_word_frequency.py --txt ./user-data/paper.txt --output ./user-data/word_freq.csv python scripts/extract_references.py --txt ./user-data/paper.txt --output ./user-data/references.csv
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 16,291 | 15,936 | -2% | 1 | 1 | 0% | 2,686 | 3,592 | +34% | 0 | 0 | — |
case-12 | fail→pass | 19,820 | 2,313 | -88% | 1 | 1 | 0% | 4,659 | 1,888 | -59% | 0 | 0 | — |
case-01 | fail→fail | 21,814 | 8,124 | -63% | 1 | 1 | 0% | 4,370 | 2,200 | -50% | 0 | 0 | — |
case-02 | fail→pass | 15,773 | 14,182 | -10% | 1 | 1 | 0% | 3,086 | 2,999 | -3% | 0 | 0 | — |
case-03 | fail→fail | 37,761 | 10,360 | -73% | 1 | 1 | 0% | 6,189 | 2,478 | -60% | 0 | 0 | — |
case-04 | pass→pass | 18,168 | 14,872 | -18% | 1 | 1 | 0% | 2,417 | 3,679 | +52% | 0 | 0 | — |
case-05 | pass→pass | 14,550 | 11,897 | -18% | 1 | 1 | 0% | 2,116 | 3,155 | +49% | 0 | 0 | — |
case-07 | fail→pass | 6,250 | 3,795 | -39% | 1 | 1 | 0% | 1,218 | 2,235 | +83% | 0 | 0 | — |
case-08 | fail→pass | 3,567 | 2,000 | -44% | 1 | 1 | 0% | 659 | 1,799 | +173% | 0 | 0 | — |
case-09 | fail→fail | 8,905 | 2,048 | -77% | 1 | 1 | 0% | 1,341 | 1,834 | +37% | 0 | 0 | — |
case-10 | fail→pass | 11,010 | 5,848 | -47% | 1 | 1 | 0% | 2,162 | 2,534 | +17% | 0 | 0 | — |
case-11 | fail→pass | 13,110 | 3,043 | -77% | 1 | 1 | 0% | 2,727 | 1,833 | -33% | 0 | 0 | — |
case-13 | fail→fail | 7,319 | 3,543 | -52% | 1 | 1 | 0% | 1,275 | 2,031 | +59% | 0 | 0 | — |
case-14 | fail→pass | 7,078 | 2,919 | -59% | 1 | 1 | 0% | 1,123 | 1,975 | +76% | 0 | 0 | — |
case-15 | fail→pass | 6,800 | 1,692 | -75% | 1 | 1 | 0% | 1,226 | 1,783 | +45% | 0 | 0 | — |
case-16 | fail→pass | 8,549 | 2,876 | -66% | 1 | 1 | 0% | 1,687 | 1,985 | +18% | 0 | 0 | — |
case-17 | fail→pass | 9,292 | 3,229 | -65% | 1 | 1 | 0% | 1,836 | 2,033 | +11% | 0 | 0 | — |
case-18 | fail→pass | 8,927 | 1,712 | -81% | 1 | 1 | 0% | 1,408 | 1,732 | +23% | 0 | 0 | — |
case-19 | pass→pass | 9,654 | 4,112 | -57% | 1 | 1 | 0% | 1,817 | 2,225 | +22% | 0 | 0 | — |
case-20 | fail→pass | 6,026 | 2,549 | -58% | 1 | 1 | 0% | 1,422 | 2,012 | +41% | 0 | 0 | — |
case-21 | fail→pass | 8,838 | 2,364 | -73% | 1 | 1 | 0% | 1,233 | 1,934 | +57% | 0 | 0 | — |
case-22 | pass→pass | 8,848 | 1,419 | -84% | 1 | 1 | 0% | 1,179 | 1,686 | +43% | 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 20 counted toward the lift figure. The other 2 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 +59 percentage points is the difference between those two pass rates over the 20 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.