Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Flyvbjerg critical case methodology: select most-likely and least-likely cases to maximize inferential power.
.claude/skills/yogsoth-ai-critical-case-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -16% | 0% |
| Size | Critical cases designed | Evaluation depth | |---|---|---| | S | 2 (one most-likely, one least-likely) | Surface | | M | 4 (two each) | Moderate | | L | 6 (three each) | Deep |
parameter-space-mapping to identify conditionsextreme-value-generation for favorable/unfavorable extremesbreakpoint-detection to test claim at critical casescontradiction-detection to evaluate outcomes<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | boundary-probing | Map parameter space, generate extreme values, test at boundaries, detect breakpoints, synthesize validity envelope. | | counterexample-heuristics | Generate counterexamples (monsters), attempt monster-barring, incorporate surviving counterexamples as lemma refinements (Lakatos method). |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | breakpoint-detection | Test a claim at extreme parameter values and detect the precise point where it breaks down. | | contradiction-detection | Evaluate whether a derivation chain has reached a genuine contradiction, absurdity, or inconclusive state. | | extreme-value-generation | Generate boundary and extreme test values for a given parameter dimension to stress-test claims. | | parameter-space-mapping | Identify all parameter dimensions along which a claim's validity might vary. |
<!-- 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 | 21,884 | 19,989 | -9% | 1 | 1 | 0% | 3,468 | 3,846 | +11% | 0 | 0 | — |
case-02 | fail→pass | 8,053 | 2,916 | -64% | 1 | 1 | 0% | 1,223 | 979 | -20% | 0 | 0 | — |
case-03 | fail→pass | 18,472 | 5,365 | -71% | 1 | 1 | 0% | 2,993 | 1,332 | -55% | 0 | 0 | — |
case-04 | fail→pass | 25,544 | 12,882 | -50% | 1 | 1 | 0% | 3,946 | 2,573 | -35% | 0 | 0 | — |
case-05 | pass→pass | 10,243 | 8,100 | -21% | 1 | 1 | 0% | 1,520 | 1,813 | +19% | 0 | 0 | — |
case-18 | fail→pass | 6,599 | 2,327 | -65% | 1 | 1 | 0% | 1,052 | 888 | -16% | 0 | 0 | — |
case-06 | pass→pass | 14,318 | 14,575 | +2% | 1 | 1 | 0% | 2,283 | 2,598 | +14% | 0 | 0 | — |
case-07 | pass→pass | 15,093 | 2,773 | -82% | 1 | 1 | 0% | 1,491 | 909 | -39% | 0 | 0 | — |
case-08 | fail→pass | 15,441 | 2,543 | -84% | 1 | 1 | 0% | 995 | 906 | -9% | 0 | 0 | — |
case-09 | fail→pass | 15,908 | 2,302 | -86% | 1 | 1 | 0% | 1,224 | 940 | -23% | 0 | 0 | — |
case-10 | fail→pass | 8,203 | 1,710 | -79% | 1 | 1 | 0% | 1,364 | 773 | -43% | 0 | 0 | — |
case-11 | fail→pass | 14,190 | 4,304 | -70% | 1 | 1 | 0% | 2,287 | 1,111 | -51% | 0 | 0 | — |
case-12 | fail→pass | 10,358 | 2,646 | -74% | 1 | 1 | 0% | 1,692 | 915 | -46% | 0 | 0 | — |
case-13 | pass→pass | 14,099 | 5,119 | -64% | 1 | 1 | 0% | 2,194 | 1,231 | -44% | 0 | 0 | — |
case-14 | pass→pass | 16,260 | 4,607 | -72% | 1 | 1 | 0% | 2,597 | 1,220 | -53% | 0 | 0 | — |
case-15 | pass→pass | 21,455 | 6,219 | -71% | 1 | 1 | 0% | 1,515 | 1,448 | -4% | 0 | 0 | — |
case-16 | fail→fail | 10,333 | 2,937 | -72% | 1 | 1 | 0% | 1,553 | 899 | -42% | 0 | 0 | — |
case-17 | pass→pass | 16,071 | 11,208 | -30% | 1 | 1 | 0% | 2,631 | 2,392 | -9% | 0 | 0 | — |
case-19 | pass→pass | 17,440 | 11,871 | -32% | 1 | 1 | 0% | 2,573 | 2,262 | -12% | 0 | 0 | — |
case-20 | fail→fail | 18,774 | 19,377 | +3% | 1 | 1 | 0% | 2,944 | 3,496 | +19% | 0 | 0 | — |
case-21 | pass→fail | 16,681 | 19,124 | +15% | 1 | 1 | 0% | 2,506 | 3,250 | +30% | 0 | 0 | — |
case-22 | fail→pass | 12,499 | 10,610 | -15% | 1 | 1 | 0% | 1,868 | 2,156 | +15% | 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 21 counted toward the lift figure. The other 1 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 +45 percentage points is the difference between those two pass rates over the 21 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.