Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Strategy: Military-grade assumption testing — Key Assumptions Check, Devil's Advocacy, Team A/B analysis to expose hidden dependencies and unexamined beliefs.
.claude/skills/yogsoth-ai-stress-test-assumption-challenge/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 54% | 0% |
Military ACT (Analytic Confidence Testing): systematically surface and attack assumptions underlying the artifact.
| Parameter | S | M | L | |---|---|---|---| | Attack vectors | 5 | 12 | 20 | | Probing rounds | 3 | 6 | 10 | | Personas | 2 | 4 | 6 | | Assumption checks | 5 | 10 | 20 |
key-assumptions-check → [enumerate and rank assumptions]
→ [for each critical assumption]:
devils-advocacy (construct counter-argument)
→ probe-execution (test assumption)
→ assumption-cascade-tracer (map dependencies)
→ finding-aggregation → attack-resilience-scoring<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | assumption-cascade | Tactic: Surface assumptions, sort by dependency, attack root assumptions first, then trace cascade failures through the dependency graph. | | structured-attack-campaign | Tactic: Full attack lifecycle — threat surface enumeration, attack vector generation, systematic probing, and finding aggregation across all surfaces. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | assumption-cascade-tracer | Build assumption dependency graphs and trace cascade failures when root assumptions are invalidated. | | devils-advocacy | Construct the strongest possible counter-argument against a position, steelmanning the opposition before attacking. | | finding-aggregation | Aggregate, deduplicate, and classify findings from multiple probes into a coherent vulnerability report. | | key-assumptions-check | Military ACT: systematically enumerate all assumptions, classify by type, and evaluate evidence strength supporting each. | | probe-execution | Execute a single attack probe against an artifact, record the result with evidence and severity classification. |
<!-- 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→pass | 48,075 | 37,808 | -21% | 1 | 1 | 0% | 6,205 | 5,795 | -7% | 0 | 0 | — |
case-02 | fail→pass | 22,592 | 31,359 | +39% | 1 | 1 | 0% | 2,188 | 3,616 | +65% | 0 | 0 | — |
case-03 | pass→pass | 55,486 | 47,754 | -14% | 1 | 1 | 0% | 8,274 | 7,391 | -11% | 0 | 0 | — |
case-04 | pass→fail | 24,365 | 26,269 | +8% | 1 | 1 | 0% | 3,308 | 5,098 | +54% | 0 | 0 | — |
case-05 | pass→fail | 39,889 | 50,902 | +28% | 1 | 1 | 0% | 7,635 | 8,851 | +16% | 0 | 0 | — |
case-06 | pass→fail | 25,473 | 45,861 | +80% | 1 | 1 | 0% | 3,449 | 7,599 | +120% | 0 | 0 | — |
case-07 | fail→pass | 22,215 | 23,244 | +5% | 1 | 1 | 0% | 2,698 | 2,358 | -13% | 0 | 0 | — |
case-08 | fail→pass | 22,857 | 18,458 | -19% | 1 | 1 | 0% | 2,676 | 2,919 | +9% | 0 | 0 | — |
case-09 | fail→pass | 19,405 | 26,042 | +34% | 1 | 1 | 0% | 2,071 | 2,601 | +26% | 0 | 0 | — |
case-10 | pass→pass | 15,856 | 15,969 | +1% | 1 | 1 | 0% | 2,497 | 3,027 | +21% | 0 | 0 | — |
case-11 | pass→pass | 14,241 | 16,012 | +12% | 1 | 1 | 0% | 2,119 | 2,384 | +13% | 0 | 0 | — |
case-12 | pass→pass | 14,920 | 14,811 | -1% | 1 | 1 | 0% | 1,631 | 2,174 | +33% | 0 | 0 | — |
case-13 | pass→pass | 12,133 | 14,658 | +21% | 1 | 1 | 0% | 1,880 | 2,177 | +16% | 0 | 0 | — |
case-14 | pass→pass | 24,259 | 22,852 | -6% | 1 | 1 | 0% | 2,950 | 3,333 | +13% | 0 | 0 | — |
case-15 | pass→pass | 19,439 | 33,820 | +74% | 1 | 1 | 0% | 2,188 | 3,314 | +51% | 0 | 0 | — |
case-16 | pass→pass | 15,619 | 14,343 | -8% | 1 | 1 | 0% | 2,344 | 2,801 | +19% | 0 | 0 | — |
case-17 | pass→pass | 16,626 | 14,499 | -13% | 1 | 1 | 0% | 1,678 | 2,097 | +25% | 0 | 0 | — |
case-18 | pass→pass | 20,925 | 23,557 | +13% | 1 | 1 | 0% | 2,601 | 3,771 | +45% | 0 | 0 | — |
case-19 | fail→fail | 15,701 | 9,601 | -39% | 1 | 1 | 0% | 1,575 | 1,332 | -15% | 0 | 0 | — |
case-20 | fail→fail | 16,190 | 14,569 | -10% | 1 | 1 | 0% | 1,794 | 2,230 | +24% | 0 | 0 | — |
case-21 | fail→fail | 8,770 | 11,358 | +30% | 1 | 1 | 0% | 1,405 | 1,606 | +14% | 0 | 0 | — |
case-22 | fail→fail | 12,998 | 19,984 | +54% | 1 | 1 | 0% | 2,093 | 3,145 | +50% | 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 +5 percentage points is the difference between those two pass rates over the 22 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.