Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Uncertainty cascade propagation — assign input distributions, sample via Monte Carlo, propagate through model, analyze output distribution, identify critical paths. Maps how input uncertainty flows to output uncertainty.
.claude/skills/yogsoth-ai-uncertainty-cascade/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-21 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 252% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 174% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 77% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 158% | 0% |
Map how input uncertainty flows through to output uncertainty.
distribution-assignment → monte-carlo-sampling → critical-path-identification
Subagent: distribution-assignment, monte-carlo-sampling, critical-path-identification Import: paper-search
Assign distributions to all uncertain inputs (use literature for informed priors), sample (Latin Hypercube for efficiency), propagate, analyze output distribution shape and spread, identify which inputs contribute most to output variance (critical path).
Focus on: which inputs, if resolved, would most reduce output uncertainty? This directly informs research prioritization.
<HARD-GATE>
- Input distributions assigned: >= 4
- Propagation completed: yes
- Output distribution characterized: yes
- Critical path identified: >= 1 path with ranked inputs
</HARD-GATE><!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | critical-path-identification | Identify which input uncertainties contribute most to output uncertainty and compute EVPI for research prioritization. | | deep-insight-paper-search | AI-powered paper summary and search. Import of literature-engine/literature-search skill. AI summary level — cite as "AI-extracted" not "paper states". | | distribution-assignment | Assign probability distributions to uncertain parameters based on available evidence and domain knowledge. | | monte-carlo-sampling | Design and execute Monte Carlo sampling strategy for uncertainty propagation through a model. |
<!-- 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 | pass→fail | 47,210 | 35,385 | -25% | 1 | 1 | 0% | 8,291 | 1,714 | -79% | 0 | 0 | — |
case-19 | pass→pass | 27,221 | 39,828 | +46% | 1 | 1 | 0% | 3,568 | 6,669 | +87% | 0 | 0 | — |
case-20 | pass→pass | 22,340 | 40,310 | +80% | 1 | 1 | 0% | 2,911 | 8,269 | +184% | 0 | 0 | — |
case-21 | fail→pass | 12,883 | 5,822 | -55% | 1 | 1 | 0% | 2,214 | 1,382 | -38% | 0 | 0 | — |
case-22 | pass→pass | 20,015 | 39,176 | +96% | 1 | 1 | 0% | 2,392 | 6,748 | +182% | 0 | 0 | — |
case-02 | pass→fail | 19,897 | 48,767 | +145% | 1 | 1 | 0% | 2,875 | 8,617 | +200% | 0 | 0 | — |
case-03 | fail→pass | 20,951 | 40,857 | +95% | 1 | 1 | 0% | 2,442 | 8,585 | +252% | 0 | 0 | — |
case-04 | pass→pass | 24,371 | 41,475 | +70% | 1 | 1 | 0% | 2,926 | 8,597 | +194% | 0 | 0 | — |
case-05 | fail→pass | 22,872 | 47,501 | +108% | 1 | 1 | 0% | 3,142 | 8,599 | +174% | 0 | 0 | — |
case-23 | pass→pass | 16,698 | 43,873 | +163% | 1 | 1 | 0% | 1,772 | 7,259 | +310% | 0 | 0 | — |
case-06 | pass→pass | 13,943 | 43,161 | +210% | 1 | 1 | 0% | 1,981 | 7,413 | +274% | 0 | 0 | — |
case-07 | pass→pass | 17,300 | 20,238 | +17% | 1 | 1 | 0% | 2,689 | 4,824 | +79% | 0 | 0 | — |
case-08 | pass→pass | 8,887 | 12,587 | +42% | 1 | 1 | 0% | 1,349 | 1,622 | +20% | 0 | 0 | — |
case-09 | pass→fail | 14,263 | 35,277 | +147% | 1 | 1 | 0% | 2,349 | 1,078 | -54% | 0 | 0 | — |
case-10 | pass→pass | 19,562 | 33,224 | +70% | 1 | 1 | 0% | 2,440 | 5,625 | +131% | 0 | 0 | — |
case-11 | pass→pass | 18,046 | 43,662 | +142% | 1 | 1 | 0% | 3,096 | 8,246 | +166% | 0 | 0 | — |
case-12 | pass→pass | 24,593 | 56,621 | +130% | 1 | 1 | 0% | 3,123 | 8,581 | +175% | 0 | 0 | — |
case-13 | fail→pass | 24,074 | 33,191 | +38% | 1 | 1 | 0% | 3,032 | 5,363 | +77% | 0 | 0 | — |
case-14 | pass→pass | 14,420 | 46,314 | +221% | 1 | 1 | 0% | 2,320 | 8,354 | +260% | 0 | 0 | — |
case-15 | fail→pass | 17,070 | 39,259 | +130% | 1 | 1 | 0% | 2,821 | 7,286 | +158% | 0 | 0 | — |
case-16 | pass→pass | 8,489 | 35,927 | +323% | 1 | 1 | 0% | 1,336 | 6,395 | +379% | 0 | 0 | — |
case-17 | fail→pass | 16,385 | 18,031 | +10% | 1 | 1 | 0% | 1,905 | 2,475 | +30% | 0 | 0 | — |
case-18 | fail→pass | 23,736 | 46,080 | +94% | 1 | 1 | 0% | 4,268 | 8,589 | +101% | 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, and 21 counted toward the lift figure. The other 2 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 +17 percentage points is the difference between those two pass rates over the 21 comparable cases. 3 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.