Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tactic: Evidence gathering, cross-examination, and quality judgment. External evidence is collected, presented, challenged, and scored for relevance and reliability.
.claude/skills/yogsoth-ai-evidence-tournament/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -3% | 0% |
Structured evidence competition — gather, present, cross-examine, and judge evidence quality.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | cross-examination | Probes defender responses for inconsistencies, logical gaps, and unsupported claims. Acts as follow-up interrogation after initial defense. | | debate-critic | Generates structured criticism from attack stance using Toulmin model. Produces claims, grounds, warrants, and rebuttals targeting artifact weaknesses. | | debate-defender | Responds to attacks with counter-evidence and counter-arguments. Defends artifact using evidence, clarification, and rebuttal while acknowledging valid criticisms. | | debate-judge | Evaluates debate exchanges, adjudicates argument quality, and produces round verdicts with confidence scores and reasoning. | | evidence-scout | Searches for external evidence supporting or opposing specific claims. Returns structured evidence with source assessment and relevance scoring. |
<!-- 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→pass | 38,954 | 37,291 | -4% | 1 | 1 | 0% | 6,219 | 5,660 | -9% | 0 | 0 | — |
case-02 | fail→pass | 38,824 | 26,775 | -31% | 1 | 1 | 0% | 6,205 | 4,920 | -21% | 0 | 0 | — |
case-03 | fail→pass | 24,382 | 25,994 | +7% | 1 | 1 | 0% | 3,879 | 4,644 | +20% | 0 | 0 | — |
case-04 | fail→pass | 11,803 | 2,591 | -78% | 1 | 1 | 0% | 1,800 | 998 | -45% | 0 | 0 | — |
case-05 | fail→pass | 14,640 | 9,342 | -36% | 1 | 1 | 0% | 2,256 | 2,190 | -3% | 0 | 0 | — |
case-06 | fail→pass | 16,203 | 3,533 | -78% | 1 | 1 | 0% | 2,510 | 1,171 | -53% | 0 | 0 | — |
case-07 | fail→pass | 11,882 | 30,475 | +156% | 1 | 1 | 0% | 1,787 | 5,615 | +214% | 0 | 0 | — |
case-08 | fail→pass | 19,995 | 17,702 | -11% | 1 | 1 | 0% | 2,864 | 3,144 | +10% | 0 | 0 | — |
case-09 | fail→pass | 18,847 | 15,075 | -20% | 1 | 1 | 0% | 2,943 | 2,868 | -3% | 0 | 0 | — |
case-10 | fail→pass | 15,359 | 3,172 | -79% | 1 | 1 | 0% | 2,251 | 1,072 | -52% | 0 | 0 | — |
case-19 | pass→pass | 14,724 | 3,200 | -78% | 1 | 1 | 0% | 2,227 | 1,147 | -48% | 0 | 0 | — |
case-11 | pass→pass | 18,579 | 6,847 | -63% | 1 | 1 | 0% | 2,698 | 1,655 | -39% | 0 | 0 | — |
case-12 | fail→pass | 13,816 | 9,950 | -28% | 1 | 1 | 0% | 1,960 | 2,048 | +4% | 0 | 0 | — |
case-13 | fail→fail | 12,436 | 7,825 | -37% | 1 | 1 | 0% | 1,900 | 1,802 | -5% | 0 | 0 | — |
case-14 | fail→fail | 13,406 | 8,213 | -39% | 1 | 1 | 0% | 2,208 | 1,976 | -11% | 0 | 0 | — |
case-15 | fail→pass | 16,961 | 14,163 | -16% | 1 | 1 | 0% | 2,562 | 2,643 | +3% | 0 | 0 | — |
case-16 | fail→pass | 16,552 | 8,843 | -47% | 1 | 1 | 0% | 2,421 | 1,971 | -19% | 0 | 0 | — |
case-17 | fail→pass | 16,580 | 14,558 | -12% | 1 | 1 | 0% | 2,576 | 2,916 | +13% | 0 | 0 | — |
case-18 | fail→pass | 15,354 | 12,046 | -22% | 1 | 1 | 0% | 2,340 | 2,432 | +4% | 0 | 0 | — |
case-20 | pass→pass | 7,353 | 7,358 | +0% | 1 | 1 | 0% | 1,043 | 1,635 | +57% | 0 | 0 | — |
case-21 | pass→pass | 12,879 | 16,023 | +24% | 1 | 1 | 0% | 2,152 | 3,125 | +45% | 0 | 0 | — |
case-22 | pass→pass | 10,139 | 8,391 | -17% | 1 | 1 | 0% | 1,580 | 1,822 | +15% | 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. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 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.