Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Identify which uncertainties would actually change the research direction decision. Compute EVPI to prioritize uncertainty reduction.
.claude/skills/yogsoth-ai-decision-sensitivity/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 149% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 31% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -56% | 0% |
Focus on uncertainties that change decisions, not just numbers.
| Base SOP | Target | ±10% Range | |----------|--------|------------| | web-search | 25 | 22–28 | | web-research | 10 | 9–11 | | paper-overview | 25 | 22–28 | | paper-search | 15 | 13–17 | | paper-research | 10 | 9–11 |
<HARD-GATE>
| SOP | Done | Target | % |
|-----|------|--------|---|
| web-search | ? | 25 | ? |
| web-research | ? | 10 | ? |
| paper-overview | ? | 25 | ? |
| paper-search | ? | 15 | ? |
| paper-research | ? | 10 | ? |
Budget Gate: OPEN/CLOSED (>=80% required to exit)
</HARD-GATE>Import: web-search, web-research, paper-overview, paper-search, paper-research Subagent: critical-path-identification, sensitivity-synthesis
Identify which uncertainties would actually change the research direction decision. Compute Expected Value of Perfect Information (EVPI) for each uncertain parameter. Focus resources on reducing uncertainty with highest EVPI.
<!-- 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. | | sensitivity-synthesis | Synthesize all sensitivity analysis results into a coherent report with prioritized recommendations. |
<!-- 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-10 | fail→pass | 13,467 | 2,403 | -82% | 1 | 1 | 0% | 824 | 799 | -3% | 0 | 0 | — |
case-01 | fail→fail | 26,069 | 5,910 | -77% | 1 | 1 | 0% | 4,422 | 1,275 | -71% | 0 | 0 | — |
case-02 | fail→pass | 16,608 | 33,584 | +102% | 1 | 1 | 0% | 2,664 | 6,631 | +149% | 0 | 0 | — |
case-03 | fail→fail | 21,835 | 38,547 | +77% | 1 | 1 | 0% | 3,441 | 6,630 | +93% | 0 | 0 | — |
case-04 | pass→fail | 25,553 | 25,245 | -1% | 1 | 1 | 0% | 2,396 | 5,508 | +130% | 0 | 0 | — |
case-11 | fail→pass | 7,667 | 4,421 | -42% | 1 | 1 | 0% | 1,663 | 1,430 | -14% | 0 | 0 | — |
case-05 | pass→pass | 28,785 | 53,618 | +86% | 1 | 1 | 0% | 5,326 | 6,613 | +24% | 0 | 0 | — |
case-06 | fail→fail | 5,145 | 7,825 | +52% | 1 | 1 | 0% | 676 | 1,734 | +157% | 0 | 0 | — |
case-07 | pass→fail | 28,184 | 9,586 | -66% | 1 | 1 | 0% | 4,926 | 1,131 | -77% | 0 | 0 | — |
case-08 | fail→pass | 28,256 | 89,884 | +218% | 1 | 1 | 0% | 5,039 | 6,622 | +31% | 0 | 0 | — |
case-09 | fail→pass | 13,017 | 2,536 | -81% | 1 | 1 | 0% | 2,052 | 894 | -56% | 0 | 0 | — |
case-12 | fail→pass | 13,623 | 3,874 | -72% | 1 | 1 | 0% | 2,145 | 1,231 | -43% | 0 | 0 | — |
case-13 | fail→pass | 7,436 | 2,956 | -60% | 1 | 1 | 0% | 1,147 | 900 | -22% | 0 | 0 | — |
case-14 | pass→pass | 10,745 | 29,915 | +178% | 1 | 1 | 0% | 2,350 | 6,658 | +183% | 0 | 0 | — |
case-15 | pass→pass | 8,357 | 13,653 | +63% | 1 | 1 | 0% | 1,231 | 2,700 | +119% | 0 | 0 | — |
case-16 | fail→pass | 10,286 | 2,157 | -79% | 1 | 1 | 0% | 639 | 784 | +23% | 0 | 0 | — |
case-17 | fail→pass | 2,399 | 2,050 | -15% | 1 | 1 | 0% | 299 | 759 | +154% | 0 | 0 | — |
case-18 | fail→pass | 3,732 | 1,966 | -47% | 1 | 1 | 0% | 503 | 746 | +48% | 0 | 0 | — |
case-19 | pass→pass | 5,029 | 10,541 | +110% | 1 | 1 | 0% | 833 | 2,399 | +188% | 0 | 0 | — |
case-24 | fail→pass | 10,292 | 2,109 | -80% | 1 | 1 | 0% | 1,443 | 771 | -47% | 0 | 0 | — |
case-20 | fail→fail | 7,892 | 6,764 | -14% | 1 | 1 | 0% | 1,576 | 1,865 | +18% | 0 | 0 | — |
case-21 | fail→pass | 28,286 | 16,379 | -42% | 1 | 1 | 0% | 5,432 | 2,105 | -61% | 0 | 0 | — |
case-22 | pass→fail | 15,475 | 7,900 | -49% | 1 | 1 | 0% | 2,273 | 1,519 | -33% | 0 | 0 | — |
case-23 | pass→pass | 10,798 | 16,341 | +51% | 1 | 1 | 0% | 1,696 | 3,368 | +99% | 0 | 0 | — |
case-25 | fail→pass | 16,299 | 3,966 | -76% | 1 | 1 | 0% | 2,531 | 1,092 | -57% | 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. 25 cases were attempted, and 23 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 +40 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.