Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Wimsatt-style multi-method cross-validation — enumerate assumptions, generate alternative models, compare results, flag divergences.
.claude/skills/yogsoth-ai-multi-model-convergence/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 149% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 274% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 124% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 157% | 0% |
Test robustness by checking if conclusions survive across different modeling choices.
Subagent: assumption-enumeration, alternative-model-generation, convergence-assessment, fragility-flagging Import: paper-research
For each key assumption, generate at least one alternative model. Run all models, compare outputs. Results that converge are robust; results that diverge are fragile.
<HARD-GATE>
- assumptions enumerated: >= 5
- alternative models generated: >= 3
- convergence assessments: >= 1
- fragility flags: assessed
</HARD-GATE><!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | alternative-model-generation | Generate alternative model formulations by relaxing, replacing, or generalizing specific assumptions. | | convergence-assessment | Compare results across multiple model variants — quantitative agreement metrics and qualitative conclusion stability. | | deep-insight-assumption-enumeration | Systematically identify all assumptions in a method/model — structural, parametric, distributional, and scope assumptions. | | deep-insight-paper-research | Full-text paper reading via three-pass Keshav method. Import of literature-engine/literature-research skill. Authoritative source for claims about paper content. | | fragility-flagging | Identify which specific assumption changes cause conclusion divergence. Rates fragility severity and plausibility of alternatives. |
<!-- 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-14 | fail→fail | 19,086 | 33,673 | +76% | 1 | 1 | 0% | 2,918 | 5,997 | +106% | 0 | 0 | — |
case-01 | fail→pass | 44,593 | 42,317 | -5% | 1 | 1 | 0% | 8,262 | 6,669 | -19% | 0 | 0 | — |
case-02 | fail→pass | 15,711 | 25,689 | +64% | 1 | 1 | 0% | 1,585 | 3,944 | +149% | 0 | 0 | — |
case-03 | fail→pass | 9,447 | 35,862 | +280% | 1 | 1 | 0% | 1,550 | 5,792 | +274% | 0 | 0 | — |
case-04 | fail→pass | 27,691 | 47,561 | +72% | 1 | 1 | 0% | 3,842 | 8,602 | +124% | 0 | 0 | — |
case-05 | pass→pass | 18,580 | 45,377 | +144% | 1 | 1 | 0% | 2,273 | 8,040 | +254% | 0 | 0 | — |
case-06 | fail→pass | 25,970 | 90,256 | +248% | 1 | 1 | 0% | 3,353 | 8,604 | +157% | 0 | 0 | — |
case-07 | fail→pass | 18,358 | 43,203 | +135% | 1 | 1 | 0% | 2,101 | 7,198 | +243% | 0 | 0 | — |
case-08 | fail→pass | 28,825 | 32,828 | +14% | 1 | 1 | 0% | 2,957 | 6,161 | +108% | 0 | 0 | — |
case-09 | fail→fail | 66,928 | 43,913 | -34% | 1 | 1 | 0% | 4,372 | 8,175 | +87% | 0 | 0 | — |
case-10 | fail→pass | 33,917 | 37,912 | +12% | 1 | 1 | 0% | 2,849 | 8,600 | +202% | 0 | 0 | — |
case-11 | fail→pass | 27,682 | 50,315 | +82% | 1 | 1 | 0% | 3,933 | 8,590 | +118% | 0 | 0 | — |
case-12 | fail→pass | 42,277 | 52,600 | +24% | 1 | 1 | 0% | 7,666 | 8,591 | +12% | 0 | 0 | — |
case-13 | fail→pass | 14,021 | 49,904 | +256% | 1 | 1 | 0% | 2,400 | 8,594 | +258% | 0 | 0 | — |
case-15 | fail→pass | 21,752 | 35,993 | +65% | 1 | 1 | 0% | 3,783 | 7,538 | +99% | 0 | 0 | — |
case-16 | pass→pass | 21,281 | 40,103 | +88% | 1 | 1 | 0% | 3,353 | 7,460 | +122% | 0 | 0 | — |
case-17 | fail→pass | 6,367 | 46,890 | +636% | 1 | 1 | 0% | 971 | 8,592 | +785% | 0 | 0 | — |
case-18 | fail→pass | 19,923 | 30,539 | +53% | 1 | 1 | 0% | 3,814 | 5,702 | +50% | 0 | 0 | — |
case-19 | fail→fail | 15,935 | 62,487 | +292% | 1 | 1 | 0% | 2,740 | 7,615 | +178% | 0 | 0 | — |
case-20 | pass→fail | 17,665 | 39,263 | +122% | 1 | 1 | 0% | 3,556 | 7,934 | +123% | 0 | 0 | — |
case-21 | pass→fail | 18,345 | 37,869 | +106% | 1 | 1 | 0% | 3,676 | 7,731 | +110% | 0 | 0 | — |
case-22 | pass→fail | 11,424 | 45,487 | +298% | 1 | 1 | 0% | 1,893 | 4,877 | +158% | 0 | 0 | — |
case-23 | pass→fail | 17,612 | 37,255 | +112% | 1 | 1 | 0% | 3,201 | 7,366 | +130% | 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. The headline lift of +43 percentage points is the difference between those two pass rates over the 23 comparable cases. 4 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.