Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Systematically collect performance data from platforms and papers
.claude/skills/yogsoth-ai-leaderboard-harvesting/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 315% | 0% |
Harvest structured performance data from leaderboard platforms (Papers With Code, benchmark-specific sites), survey papers, and official benchmark repositories. Produces deduplicated, provenance-tracked score collections.
Identify and scrape all relevant leaderboard sources:
Yield: List of leaderboard URLs + initial method counts per source.
For methods not covered by leaderboards, extract scores directly from papers:
Yield: Raw score tuples with paper provenance.
Compare scores across sources for the same method-dataset-metric triple:
Yield: Validated score set with confidence annotations.
Consolidate all sources into a single canonical dataset:
Yield: Unified performance dataset ready for analysis.
| Metric | Floor | |--------|-------| | Leaderboard sources checked | 3 | | Methods with scores | 15 | | Cross-validated score pairs | 10 | | Deduplication conflicts resolved | 5 |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | discrepancy-identification | Compare same-method scores across sources, flag significant deviations | | method-discovery | Identify all relevant methods via literature, leaderboards, citation chains | | score-extraction | Extract (Task, Dataset, Metric, Score, Conditions) tuples from a paper |
<!-- 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-22 | pass→pass | 19,735 | 22,248 | +13% | 1 | 1 | 0% | 3,696 | 4,696 | +27% | 0 | 0 | — |
case-02 | fail→fail | 32,619 | 45,686 | +40% | 1 | 1 | 0% | 7,422 | 7,788 | +5% | 0 | 0 | — |
case-03 | fail→pass | 25,828 | 23,796 | -8% | 1 | 1 | 0% | 4,444 | 5,033 | +13% | 0 | 0 | — |
case-01 | fail→pass | 83,917 | 38,127 | -55% | 1 | 1 | 0% | 8,274 | 8,387 | +1% | 0 | 0 | — |
case-04 | fail→fail | 19,882 | 16,779 | -16% | 1 | 1 | 0% | 2,247 | 2,185 | -3% | 0 | 0 | — |
case-05 | pass→pass | 25,237 | 23,767 | -6% | 1 | 1 | 0% | 3,764 | 5,516 | +47% | 0 | 0 | — |
case-06 | fail→pass | 26,198 | 19,420 | -26% | 1 | 1 | 0% | 3,558 | 2,185 | -39% | 0 | 0 | — |
case-07 | fail→pass | 18,779 | 8,648 | -54% | 1 | 1 | 0% | 2,253 | 1,965 | -13% | 0 | 0 | — |
case-08 | fail→pass | 10,751 | 32,827 | +205% | 1 | 1 | 0% | 1,514 | 6,290 | +315% | 0 | 0 | — |
case-09 | pass→pass | 14,343 | 9,497 | -34% | 1 | 1 | 0% | 2,507 | 2,470 | -1% | 0 | 0 | — |
case-10 | fail→fail | 17,615 | 7,767 | -56% | 1 | 1 | 0% | 3,124 | 892 | -71% | 0 | 0 | — |
case-11 | pass→pass | 20,305 | 15,951 | -21% | 1 | 1 | 0% | 2,605 | 2,751 | +6% | 0 | 0 | — |
case-16 | fail→pass | 19,269 | 18,870 | -2% | 1 | 1 | 0% | 3,577 | 3,938 | +10% | 0 | 0 | — |
case-12 | pass→pass | 16,866 | 21,095 | +25% | 1 | 1 | 0% | 2,954 | 3,112 | +5% | 0 | 0 | — |
case-13 | fail→fail | 32,837 | 4,380 | -87% | 1 | 1 | 0% | 1,007 | 1,089 | +8% | 0 | 0 | — |
case-14 | fail→pass | 27,900 | 12,736 | -54% | 1 | 1 | 0% | 3,446 | 1,660 | -52% | 0 | 0 | — |
case-15 | pass→pass | 14,191 | 6,359 | -55% | 1 | 1 | 0% | 2,572 | 1,305 | -49% | 0 | 0 | — |
case-17 | fail→fail | 12,844 | 15,826 | +23% | 1 | 1 | 0% | 2,317 | 1,906 | -18% | 0 | 0 | — |
case-18 | pass→pass | 17,588 | 14,555 | -17% | 1 | 1 | 0% | 2,271 | 2,619 | +15% | 0 | 0 | — |
case-19 | fail→pass | 16,271 | 19,232 | +18% | 1 | 1 | 0% | 2,465 | 1,754 | -29% | 0 | 0 | — |
case-20 | pass→pass | 17,543 | 16,049 | -9% | 1 | 1 | 0% | 3,487 | 3,901 | +12% | 0 | 0 | — |
case-21 | pass→pass | 24,167 | 27,174 | +12% | 1 | 1 | 0% | 4,887 | 5,251 | +7% | 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 21 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 +36 percentage points is the difference between those two pass rates over the 21 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.