Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Import SOP: paper AI summary reading (from literature-engine skill)
.claude/skills/yogsoth-ai-experiment-execution-paper-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -79% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -68% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -71% | 0% |
Paper AI summary reading — imported from literature-engine skill.
Import — invokes literature-engine skill's paper-search SOP.
Available to all campaigns for quickly obtaining AI-generated paper summaries (method overview, key contributions, experimental setup, etc.).
get_paper_contentpaper<!-- 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 |
<!-- 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-01 | fail→fail | 24,325 | 7,129 | -71% | 1 | 1 | 0% | 4,024 | 630 | -84% | 0 | 0 | — |
case-02 | fail→fail | 24,515 | 11,100 | -55% | 1 | 1 | 0% | 4,064 | 602 | -85% | 0 | 0 | — |
case-03 | fail→fail | 24,635 | 6,384 | -74% | 1 | 1 | 0% | 4,372 | 527 | -88% | 0 | 0 | — |
case-04 | pass→pass | 23,601 | 17,788 | -25% | 1 | 1 | 0% | 4,164 | 3,268 | -22% | 0 | 0 | — |
case-05 | pass→pass | 14,069 | 15,513 | +10% | 1 | 1 | 0% | 3,240 | 3,804 | +17% | 0 | 0 | — |
case-06 | pass→pass | 23,314 | 20,590 | -12% | 1 | 1 | 0% | 4,929 | 4,767 | -3% | 0 | 0 | — |
case-16 | fail→fail | 11,553 | 1,777 | -85% | 1 | 1 | 0% | 1,652 | 435 | -74% | 0 | 0 | — |
case-07 | pass→pass | 7,498 | 1,727 | -77% | 1 | 1 | 0% | 1,267 | 394 | -69% | 0 | 0 | — |
case-08 | fail→pass | 8,252 | 1,338 | -84% | 1 | 1 | 0% | 1,351 | 356 | -74% | 0 | 0 | — |
case-09 | pass→pass | 13,091 | 1,482 | -89% | 1 | 1 | 0% | 2,086 | 368 | -82% | 0 | 0 | — |
case-10 | fail→pass | 11,183 | 1,605 | -86% | 1 | 1 | 0% | 1,690 | 357 | -79% | 0 | 0 | — |
case-15 | fail→pass | 10,468 | 1,765 | -83% | 1 | 1 | 0% | 1,502 | 419 | -72% | 0 | 0 | — |
case-11 | pass→pass | 13,474 | 7,080 | -47% | 1 | 1 | 0% | 2,012 | 1,249 | -38% | 0 | 0 | — |
case-12 | fail→pass | 7,956 | 1,674 | -79% | 1 | 1 | 0% | 1,201 | 385 | -68% | 0 | 0 | — |
case-13 | pass→fail | 4,830 | 2,122 | -56% | 1 | 1 | 0% | 726 | 486 | -33% | 0 | 0 | — |
case-14 | pass→pass | 11,754 | 6,437 | -45% | 1 | 1 | 0% | 1,871 | 1,205 | -36% | 0 | 0 | — |
case-17 | fail→pass | 14,993 | 3,312 | -78% | 1 | 1 | 0% | 2,397 | 700 | -71% | 0 | 0 | — |
case-18 | fail→pass | 10,934 | 2,254 | -79% | 1 | 1 | 0% | 1,643 | 480 | -71% | 0 | 0 | — |
case-19 | pass→pass | 16,248 | 10,664 | -34% | 1 | 1 | 0% | 2,577 | 1,876 | -27% | 0 | 0 | — |
case-20 | pass→pass | 16,602 | 10,715 | -35% | 1 | 1 | 0% | 2,409 | 1,870 | -22% | 0 | 0 | — |
case-21 | fail→pass | 6,178 | 1,856 | -70% | 1 | 1 | 0% | 949 | 444 | -53% | 0 | 0 | — |
case-22 | fail→pass | 5,350 | 2,305 | -57% | 1 | 1 | 0% | 777 | 501 | -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 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 +32 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.