Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →10 document processing skills. Trigger: extracting text from PDFs, parsing references, document Q&A. Design: parsing pipelines (GROBID, marker) and structured extraction tools.
.claude/skills/brycewang-stanford-document-skills/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -33% | 0% |
Select the skill matching the user's need, then read its SKILL.md.
| Skill | Description | |-------|-------------| | anystyle-api | Citation reference parser using machine learning | | docsgpt-guide | Deploy DocsGPT for private document analysis and research knowledge bases | | grobid-pdf-parsing | Extract structured text, metadata, and references from academic PDFs | | large-document-reader | Split and read long documents chapter-by-chapter for structured analysis | | markdown-academic-guide | Write academic papers in Markdown with Pandoc for multi-format output | | paper-parse-guide | Deep dual-mode reading of academic papers from PDF or URL sources | | pdf-extraction-guide | PDF parsing, text extraction, and document format conversion | | zotero-addon-market-guide | Plugin marketplace and discovery platform for Zotero | | zotero-night-theme-guide | Dark mode theme plugin for Zotero reference manager | | zotero-style-guide | Feature-rich Zotero plugin for UI customization and styling |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,516 | 10,160 | -3% | 1 | 1 | 0% | 1,458 | 1,303 | -11% | 0 | 0 | — |
case-02 | pass→fail | 19,284 | 7,570 | -61% | 1 | 1 | 0% | 2,622 | 1,057 | -60% | 0 | 0 | — |
case-03 | fail→fail | 14,699 | 5,154 | -65% | 1 | 1 | 0% | 2,241 | 602 | -73% | 0 | 0 | — |
case-04 | fail→pass | 14,525 | 6,340 | -56% | 1 | 1 | 0% | 2,018 | 1,004 | -50% | 0 | 0 | — |
case-05 | fail→pass | 17,908 | 4,949 | -72% | 1 | 1 | 0% | 2,484 | 1,222 | -51% | 0 | 0 | — |
case-06 | fail→pass | 10,022 | 6,700 | -33% | 1 | 1 | 0% | 1,460 | 889 | -39% | 0 | 0 | — |
case-07 | fail→fail | 10,400 | 4,295 | -59% | 1 | 1 | 0% | 1,760 | 718 | -59% | 0 | 0 | — |
case-08 | fail→pass | 9,959 | 3,683 | -63% | 1 | 1 | 0% | 1,456 | 969 | -33% | 0 | 0 | — |
case-09 | fail→pass | 11,249 | 4,769 | -58% | 1 | 1 | 0% | 1,804 | 1,168 | -35% | 0 | 0 | — |
case-10 | fail→pass | 18,372 | 3,341 | -82% | 1 | 1 | 0% | 2,680 | 896 | -67% | 0 | 0 | — |
case-11 | fail→pass | 14,533 | 2,168 | -85% | 1 | 1 | 0% | 2,096 | 850 | -59% | 0 | 0 | — |
case-12 | fail→pass | 8,663 | 2,415 | -72% | 1 | 1 | 0% | 1,457 | 821 | -44% | 0 | 0 | — |
case-13 | fail→pass | 13,968 | 1,705 | -88% | 1 | 1 | 0% | 2,586 | 711 | -73% | 0 | 0 | — |
case-14 | fail→pass | 10,333 | 2,518 | -76% | 1 | 1 | 0% | 1,492 | 865 | -42% | 0 | 0 | — |
case-15 | fail→pass | 10,998 | 2,172 | -80% | 1 | 1 | 0% | 1,524 | 757 | -50% | 0 | 0 | — |
case-16 | fail→pass | 12,474 | 1,641 | -87% | 1 | 1 | 0% | 1,870 | 560 | -70% | 0 | 0 | — |
case-17 | fail→pass | 14,193 | 2,075 | -85% | 1 | 1 | 0% | 2,025 | 726 | -64% | 0 | 0 | — |
case-18 | pass→pass | 12,784 | 6,402 | -50% | 1 | 1 | 0% | 1,992 | 824 | -59% | 0 | 0 | — |
case-19 | fail→pass | 7,837 | 3,195 | -59% | 1 | 1 | 0% | 1,209 | 909 | -25% | 0 | 0 | — |
case-20 | pass→pass | 5,479 | 5,051 | -8% | 1 | 1 | 0% | 829 | 1,243 | +50% | 0 | 0 | — |
case-21 | pass→pass | 8,086 | 7,671 | -5% | 1 | 1 | 0% | 1,318 | 1,715 | +30% | 0 | 0 | — |
case-22 | pass→pass | 11,258 | 10,790 | -4% | 1 | 1 | 0% | 1,909 | 2,165 | +13% | 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 19 counted toward the lift figure. The other 3 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 +64 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is 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.