Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Measure how much conclusions change when each assumption is negated. Ranks assumptions by their impact on the final result.
.claude/skills/yogsoth-ai-assumption-criticality/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 263% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 164% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 231% | 0% |
| case-02 | ✓→✗ | ▼ Worse | 119% | 0% |
| case-09 | ✓→✗ | ▼ Worse | 112% | 0% |
Rank assumptions by their impact on conclusions.
| Base SOP | Target | ±10% Range | |----------|--------|------------| | web-search | 30 | 27–33 | | web-research | 10 | 9–11 | | paper-overview | 30 | 27–33 | | paper-search | 20 | 18–22 | | paper-research | 10 | 9–11 |
<HARD-GATE>
| SOP | Done | Target | % |
|-----|------|--------|---|
| web-search | ? | 30 | ? |
| web-research | ? | 10 | ? |
| paper-overview | ? | 30 | ? |
| paper-search | ? | 20 | ? |
| 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: assumption-extraction, negation-definition, re-derivation, conclusion-sensitivity-measurement
Extract assumptions, define negation for each, re-derive conclusions under negated assumption, measure how much the conclusion changes. Rank by conclusion sensitivity.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | deep-insight-assumption-perturbation | One-at-a-time assumption perturbation — extract assumptions, define negations, re-derive conclusions under each negation, measure sensitivity. Identifies which assumptions are load-bearing. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | conclusion-sensitivity-measurement | Quantify how much conclusions change across all assumption negations and produce a sensitivity ranking. | | deep-insight-assumption-extraction | Systematically extract all assumptions (stated, implicit, boundary, mathematical, practical) from a method or model. | | negation-definition | Define strongest plausible alternatives (negations) for each assumption to enable perturbation analysis. | | re-derivation | Re-derive conclusions under a negated assumption, tracking where the derivation diverges from the original. |
<!-- 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→fail | 10,649 | 10,910 | +2% | 1 | 1 | 0% | 1,682 | 1,117 | -34% | 0 | 0 | — |
case-02 | pass→fail | 20,596 | 30,516 | +48% | 1 | 1 | 0% | 3,077 | 6,747 | +119% | 0 | 0 | — |
case-03 | fail→fail | 17,969 | 36,152 | +101% | 1 | 1 | 0% | 2,833 | 6,773 | +139% | 0 | 0 | — |
case-04 | fail→fail | 14,783 | 9,709 | -34% | 1 | 1 | 0% | 2,312 | 1,326 | -43% | 0 | 0 | — |
case-05 | fail→fail | 10,572 | 37,528 | +255% | 1 | 1 | 0% | 1,505 | 5,561 | +270% | 0 | 0 | — |
case-06 | fail→fail | 16,055 | 5,560 | -65% | 1 | 1 | 0% | 2,286 | 1,283 | -44% | 0 | 0 | — |
case-07 | fail→pass | 10,376 | 32,569 | +214% | 1 | 1 | 0% | 1,569 | 5,690 | +263% | 0 | 0 | — |
case-08 | fail→pass | 16,369 | 37,934 | +132% | 1 | 1 | 0% | 2,315 | 6,105 | +164% | 0 | 0 | — |
case-09 | pass→fail | 18,787 | 33,719 | +79% | 1 | 1 | 0% | 3,184 | 6,765 | +112% | 0 | 0 | — |
case-10 | pass→fail | 18,246 | 9,243 | -49% | 1 | 1 | 0% | 2,603 | 1,171 | -55% | 0 | 0 | — |
case-11 | fail→pass | 10,014 | 25,876 | +158% | 1 | 1 | 0% | 1,473 | 4,871 | +231% | 0 | 0 | — |
case-12 | pass→pass | 19,488 | 36,326 | +86% | 1 | 1 | 0% | 3,028 | 6,769 | +124% | 0 | 0 | — |
case-13 | fail→fail | 19,554 | 34,622 | +77% | 1 | 1 | 0% | 2,997 | 6,762 | +126% | 0 | 0 | — |
case-14 | fail→fail | 19,542 | 5,874 | -70% | 1 | 1 | 0% | 2,662 | 1,375 | -48% | 0 | 0 | — |
case-15 | fail→fail | 14,634 | 34,105 | +133% | 1 | 1 | 0% | 2,155 | 6,769 | +214% | 0 | 0 | — |
case-16 | fail→fail | 19,746 | 6,854 | -65% | 1 | 1 | 0% | 2,931 | 1,576 | -46% | 0 | 0 | — |
case-17 | pass→fail | 21,909 | 11,319 | -48% | 1 | 1 | 0% | 2,994 | 1,198 | -60% | 0 | 0 | — |
case-18 | fail→fail | 18,597 | 5,866 | -68% | 1 | 1 | 0% | 2,785 | 1,435 | -48% | 0 | 0 | — |
case-19 | pass→pass | 39,482 | 47,456 | +20% | 1 | 1 | 0% | 3,062 | 5,909 | +93% | 0 | 0 | — |
case-20 | pass→fail | 21,008 | 31,617 | +50% | 1 | 1 | 0% | 4,073 | 6,773 | +66% | 0 | 0 | — |
case-21 | pass→fail | 19,942 | 29,370 | +47% | 1 | 1 | 0% | 3,327 | 5,753 | +73% | 0 | 0 | — |
case-22 | pass→fail | 17,149 | 37,507 | +119% | 1 | 1 | 0% | 2,654 | 6,767 | +155% | 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 20 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 -38 percentage points is the difference between those two pass rates over the 20 comparable cases. 10 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.