Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Paper AI summary report via alphaxiv get_paper_content. Import of literature-engine/paper-search skill. Structured AI-generated intermediate report.
.claude/skills/yogsoth-ai-stress-test-paper-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -51% | 0% |
Paper AI summary report — structured intermediate report optimized for LLM consumption.
Import — strictly follow literature-engine/paper-search skill protocol.
Returns AI-generated summary. Sufficient for methodology and results claims. For exact quotes or detailed data, use paper-research (full text).
Quantity target is set by the calling strategy's budget table. This SOP executes one unit = one get_paper_content call.
literature-engine repo → skills/paper-search/SKILL.md
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | literature-search | Medium-depth literature search — read AI-summarized reports for every paper analyzed | | stress-test-paper-research | Paper full text access via alphaxiv answer_pdf_queries or get_paper_content(fullText=true). Import of literature-engine/paper-research skill. Raw extracted text for precise claims. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→fail | 33,799 | 51,256 | +52% | 1 | 1 | 0% | 5,868 | 8,470 | +44% | 0 | 0 | — |
case-01 | fail→fail | 25,888 | 48,426 | +87% | 1 | 1 | 0% | 4,085 | 6,343 | +55% | 0 | 0 | — |
case-02 | fail→pass | 31,928 | 3,393 | -89% | 1 | 1 | 0% | 1,675 | 909 | -46% | 0 | 0 | — |
case-03 | fail→pass | 14,499 | 9,707 | -33% | 1 | 1 | 0% | 1,319 | 1,048 | -21% | 0 | 0 | — |
case-05 | fail→pass | 26,842 | 6,656 | -75% | 1 | 1 | 0% | 1,194 | 525 | -56% | 0 | 0 | — |
case-06 | fail→pass | 17,800 | 9,513 | -47% | 1 | 1 | 0% | 2,155 | 1,010 | -53% | 0 | 0 | — |
case-07 | fail→pass | 17,534 | 3,901 | -78% | 1 | 1 | 0% | 1,843 | 898 | -51% | 0 | 0 | — |
case-08 | pass→pass | 9,321 | 1,799 | -81% | 1 | 1 | 0% | 1,457 | 588 | -60% | 0 | 0 | — |
case-09 | fail→pass | 14,979 | 6,986 | -53% | 1 | 1 | 0% | 1,611 | 573 | -64% | 0 | 0 | — |
case-10 | fail→fail | 12,810 | 2,598 | -80% | 1 | 1 | 0% | 1,159 | 676 | -42% | 0 | 0 | — |
case-11 | fail→fail | 14,398 | 23,837 | +66% | 1 | 1 | 0% | 2,437 | 4,499 | +85% | 0 | 0 | — |
case-12 | pass→pass | 14,904 | 7,934 | -47% | 1 | 1 | 0% | 1,443 | 673 | -53% | 0 | 0 | — |
case-13 | fail→pass | 16,479 | 6,791 | -59% | 1 | 1 | 0% | 1,939 | 508 | -74% | 0 | 0 | — |
case-18 | fail→pass | 16,349 | 2,686 | -84% | 1 | 1 | 0% | 1,851 | 762 | -59% | 0 | 0 | — |
case-14 | fail→pass | 15,399 | 7,403 | -52% | 1 | 1 | 0% | 1,681 | 664 | -60% | 0 | 0 | — |
case-15 | pass→fail | 7,933 | 7,199 | -9% | 1 | 1 | 0% | 473 | 607 | +28% | 0 | 0 | — |
case-16 | fail→pass | 18,204 | 9,708 | -47% | 1 | 1 | 0% | 2,025 | 994 | -51% | 0 | 0 | — |
case-17 | pass→pass | 23,817 | 22,235 | -7% | 1 | 1 | 0% | 3,223 | 2,919 | -9% | 0 | 0 | — |
case-19 | fail→fail | 16,960 | 2,234 | -87% | 1 | 1 | 0% | 2,019 | 608 | -70% | 0 | 0 | — |
case-20 | fail→pass | 18,975 | 21,897 | +15% | 1 | 1 | 0% | 2,236 | 2,708 | +21% | 0 | 0 | — |
case-21 | fail→pass | 16,390 | 7,385 | -55% | 1 | 1 | 0% | 1,808 | 558 | -69% | 0 | 0 | — |
case-22 | pass→pass | 11,227 | 2,241 | -80% | 1 | 1 | 0% | 874 | 622 | -29% | 0 | 0 | — |
case-23 | fail→fail | 11,009 | 6,997 | -36% | 1 | 1 | 0% | 1,787 | 607 | -66% | 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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 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 +48 percentage points is the difference between those two pass rates over the 22 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.