Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Strategy: Williamson-style precise thought experiments — construct carefully specified counterfactual scenarios to test whether conclusions depend on contingent features.
.claude/skills/yogsoth-ai-thought-experiment/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 34% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 91% | 0% |
Williamson methodology: construct precise, well-specified counterfactual scenarios that isolate individual variables.
| Parameter | S | M | L | |---|---|---|---| | Thought experiments | 3 | 8 | 15 | | Variables isolated | 3 | 6 | 12 | | Scenario precision checks | 1 | 3 | 6 |
causal-claim-extraction → factor-enumeration
→ [for each contingent feature]:
counterfactual-scenario-construction (precise scenario)
→ flip-point-detection (does conclusion hold?)
→ necessity-evaluation (is this genuinely necessary?)
→ load-bearing-identification (essential vs contingent)<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | causal-necessity-testing | Tactic: Extract causal claims, evaluate probability of necessity (PN) and sufficiency (PS) for each, classify into necessity-sufficiency quadrants. | | minimal-change-search | Tactic: Generate candidate changes, detect flip-points where conclusion reverses, measure fragility as distance to nearest flip. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | causal-claim-extraction | Extract all causal claims (X causes Y, X leads to Y, X enables Y) from an artifact, producing a structured list of cause-effect pairs. | | counterfactual-scenario-construction | Construct precise, internally consistent counterfactual scenarios where specified factors are altered, then reason about the resulting conclusion. | | factor-enumeration | List all key factors, conditions, and assumptions that support or enable the artifact's conclusion. | | flip-point-detection | Find the minimal change magnitude along a dimension that causes the conclusion to flip from true to false. | | load-bearing-identification | Identify which factors are "load-bearing walls" — factors whose removal would collapse the conclusion. | | necessity-evaluation | Evaluate the probability of necessity (PN) for a causal factor — would the conclusion fail if this factor were absent? |
<!-- 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 | 28,792 | 31,052 | +8% | 1 | 1 | 0% | 3,737 | 5,022 | +34% | 0 | 0 | — |
case-02 | fail→fail | 28,332 | 26,368 | -7% | 1 | 1 | 0% | 3,717 | 4,058 | +9% | 0 | 0 | — |
case-03 | fail→pass | 30,289 | 39,031 | +29% | 1 | 1 | 0% | 3,941 | 5,957 | +51% | 0 | 0 | — |
case-04 | pass→pass | 20,238 | 17,195 | -15% | 1 | 1 | 0% | 2,339 | 2,700 | +15% | 0 | 0 | — |
case-05 | fail→pass | 29,879 | 40,674 | +36% | 1 | 1 | 0% | 4,056 | 6,433 | +59% | 0 | 0 | — |
case-06 | fail→pass | 37,808 | 54,636 | +45% | 1 | 1 | 0% | 5,475 | 8,941 | +63% | 0 | 0 | — |
case-07 | fail→pass | 16,766 | 29,845 | +78% | 1 | 1 | 0% | 2,734 | 5,215 | +91% | 0 | 0 | — |
case-08 | fail→pass | 24,408 | 30,177 | +24% | 1 | 1 | 0% | 4,299 | 5,642 | +31% | 0 | 0 | — |
case-09 | fail→pass | 25,565 | 21,293 | -17% | 1 | 1 | 0% | 3,263 | 4,981 | +53% | 0 | 0 | — |
case-10 | fail→pass | 5,524 | 10,597 | +92% | 1 | 1 | 0% | 806 | 2,270 | +182% | 0 | 0 | — |
case-11 | pass→pass | 26,779 | 44,249 | +65% | 1 | 1 | 0% | 3,621 | 8,939 | +147% | 0 | 0 | — |
case-12 | pass→pass | 20,680 | 46,786 | +126% | 1 | 1 | 0% | 2,393 | 7,716 | +222% | 0 | 0 | — |
case-13 | fail→pass | 17,386 | 29,216 | +68% | 1 | 1 | 0% | 2,551 | 4,558 | +79% | 0 | 0 | — |
case-14 | pass→pass | 31,968 | 35,809 | +12% | 1 | 1 | 0% | 4,088 | 6,945 | +70% | 0 | 0 | — |
case-15 | fail→fail | 19,088 | 26,112 | +37% | 1 | 1 | 0% | 2,346 | 4,279 | +82% | 0 | 0 | — |
case-16 | pass→pass | 25,927 | 41,032 | +58% | 1 | 1 | 0% | 2,925 | 5,909 | +102% | 0 | 0 | — |
case-17 | fail→pass | 21,757 | 35,387 | +63% | 1 | 1 | 0% | 2,621 | 5,692 | +117% | 0 | 0 | — |
case-18 | pass→pass | 17,361 | 23,450 | +35% | 1 | 1 | 0% | 3,032 | 4,731 | +56% | 0 | 0 | — |
case-19 | fail→fail | 19,823 | 47,468 | +139% | 1 | 1 | 0% | 3,002 | 7,030 | +134% | 0 | 0 | — |
case-20 | fail→fail | 11,379 | 45,692 | +302% | 1 | 1 | 0% | 1,091 | 7,982 | +632% | 0 | 0 | — |
case-21 | pass→pass | 13,703 | 12,949 | -6% | 1 | 1 | 0% | 1,717 | 2,774 | +62% | 0 | 0 | — |
case-22 | pass→fail | 19,757 | 32,592 | +65% | 1 | 1 | 0% | 3,070 | 6,376 | +108% | 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 +41 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.