Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Full-text paper reading via three-pass Keshav method. Import of literature-engine/literature-research skill. Authoritative source for claims about paper content.
.claude/skills/yogsoth-ai-deep-insight-paper-research/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -79% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-09 | ✓→✗ | ▼ Worse | -77% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -75% | 0% |
Full-text paper reading via three-pass Keshav method.
Import — strictly follow literature-engine/literature-research skill protocol.
Full text reading — this is the authoritative source for claims about paper methodology, results, and contributions. Findings from this SOP can be cited directly.
Quantity target is set by the calling strategy's budget table. This SOP executes one unit = one paper full-text read + structured analysis.
literature-engine repo → skills/literature-research/SKILL.md
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | literature-research | Deep literature research — raw full text reading and targeted PDF queries for rigorous analysis |
<!-- 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 | 26,524 | 5,743 | -78% | 1 | 1 | 0% | 4,506 | 492 | -89% | 0 | 0 | — |
case-02 | fail→fail | 27,906 | 35,036 | +26% | 1 | 1 | 0% | 5,105 | 6,385 | +25% | 0 | 0 | — |
case-03 | fail→pass | 16,133 | 22,811 | +41% | 1 | 1 | 0% | 3,030 | 4,111 | +36% | 0 | 0 | — |
case-04 | pass→pass | 11,969 | 2,655 | -78% | 1 | 1 | 0% | 1,858 | 664 | -64% | 0 | 0 | — |
case-05 | pass→pass | 14,515 | 11,953 | -18% | 1 | 1 | 0% | 2,335 | 2,067 | -11% | 0 | 0 | — |
case-06 | pass→pass | 13,894 | 14,182 | +2% | 1 | 1 | 0% | 2,298 | 2,661 | +16% | 0 | 0 | — |
case-07 | pass→pass | 9,578 | 10,254 | +7% | 1 | 1 | 0% | 1,483 | 1,739 | +17% | 0 | 0 | — |
case-08 | pass→pass | 14,231 | 7,929 | -44% | 1 | 1 | 0% | 2,133 | 1,365 | -36% | 0 | 0 | — |
case-09 | pass→fail | 17,126 | 2,311 | -87% | 1 | 1 | 0% | 2,612 | 593 | -77% | 0 | 0 | — |
case-10 | pass→pass | 17,752 | 13,414 | -24% | 1 | 1 | 0% | 2,695 | 2,160 | -20% | 0 | 0 | — |
case-11 | pass→pass | 2,745 | 2,733 | -0% | 1 | 1 | 0% | 442 | 579 | +31% | 0 | 0 | — |
case-12 | pass→pass | 13,022 | 9,018 | -31% | 1 | 1 | 0% | 1,953 | 1,705 | -13% | 0 | 0 | — |
case-13 | pass→pass | 11,681 | 6,668 | -43% | 1 | 1 | 0% | 1,892 | 1,222 | -35% | 0 | 0 | — |
case-14 | fail→pass | 15,498 | 1,961 | -87% | 1 | 1 | 0% | 2,469 | 507 | -79% | 0 | 0 | — |
case-15 | pass→pass | 10,862 | 6,368 | -41% | 1 | 1 | 0% | 1,660 | 1,039 | -37% | 0 | 0 | — |
case-16 | pass→pass | 8,597 | 5,695 | -34% | 1 | 1 | 0% | 1,369 | 1,131 | -17% | 0 | 0 | — |
case-17 | fail→pass | 14,482 | 2,410 | -83% | 1 | 1 | 0% | 2,239 | 619 | -72% | 0 | 0 | — |
case-18 | fail→fail | 3,427 | 2,080 | -39% | 1 | 1 | 0% | 495 | 491 | -1% | 0 | 0 | — |
case-19 | pass→pass | 11,022 | 2,029 | -82% | 1 | 1 | 0% | 1,663 | 492 | -70% | 0 | 0 | — |
case-24 | fail→fail | 1,655 | 4,161 | +151% | 1 | 1 | 0% | 253 | 791 | +213% | 0 | 0 | — |
case-20 | pass→pass | 17,765 | 18,562 | +4% | 1 | 1 | 0% | 2,827 | 3,159 | +12% | 0 | 0 | — |
case-21 | pass→fail | 13,718 | 1,988 | -86% | 1 | 1 | 0% | 2,204 | 552 | -75% | 0 | 0 | — |
case-22 | pass→pass | 7,507 | 8,528 | +14% | 1 | 1 | 0% | 1,378 | 1,667 | +21% | 0 | 0 | — |
case-23 | pass→pass | 2,833 | 2,759 | -3% | 1 | 1 | 0% | 425 | 691 | +63% | 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. 24 cases were attempted, and 23 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 +4 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 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.