Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Which assumptions are most fragile? — Vulnerability ranking + impact assessment of experiment assumptions
.claude/skills/yogsoth-ai-assumption-constraint/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 12% | 0% |
Systematic assumption vulnerability analysis:
Assumption categories: | Category | Examples | |----------|----------| | Technical | Method convergence, architecture suitability | | Data | Availability, quality, representativeness | | Resource | Sufficiency of compute, time, expertise | | Environmental | Tool stability, API access, policy | | Theoretical | Effect existence, measurability, magnitude |
assumption-challenging SOPresource-quantification SOPsensitivity-ranking tactic| Resource | Budget | Notes | |----------|--------|-------| | Subagent calls | ≤5 | 2 SOPs + synthesis | | Iterations | ≤2 | Re-rank if new assumptions surface | | Output size | ≤3000 tokens | Ranked table + validation plan |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | sensitivity-ranking | Rank constraints by sensitivity — which ones most impact the outcome if they shift |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | assumption-challenging | Challenge each assumption's validity — shared cross-repo SOP | | resource-quantification | Quantify resource demand vs supply vs gap for each resource category |
<!-- 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 | 25,287 | 7,550 | -70% | 1 | 1 | 0% | 4,029 | 971 | -76% | 0 | 0 | — |
case-02 | fail→fail | 18,206 | 25,838 | +42% | 1 | 1 | 0% | 2,671 | 4,305 | +61% | 0 | 0 | — |
case-03 | pass→pass | 19,251 | 37,380 | +94% | 1 | 1 | 0% | 3,179 | 6,328 | +99% | 0 | 0 | — |
case-04 | pass→pass | 16,742 | 27,451 | +64% | 1 | 1 | 0% | 2,566 | 4,689 | +83% | 0 | 0 | — |
case-05 | pass→pass | 12,107 | 23,497 | +94% | 1 | 1 | 0% | 2,318 | 4,569 | +97% | 0 | 0 | — |
case-06 | fail→pass | 22,726 | 37,456 | +65% | 1 | 1 | 0% | 3,902 | 5,572 | +43% | 0 | 0 | — |
case-07 | pass→pass | 19,420 | 17,156 | -12% | 1 | 1 | 0% | 3,208 | 3,143 | -2% | 0 | 0 | — |
case-08 | fail→pass | 17,186 | 11,246 | -35% | 1 | 1 | 0% | 2,798 | 1,864 | -33% | 0 | 0 | — |
case-09 | pass→pass | 19,155 | 7,152 | -63% | 1 | 1 | 0% | 2,810 | 1,633 | -42% | 0 | 0 | — |
case-10 | fail→pass | 10,399 | 2,301 | -78% | 1 | 1 | 0% | 1,613 | 925 | -43% | 0 | 0 | — |
case-11 | pass→pass | 13,352 | 2,704 | -80% | 1 | 1 | 0% | 2,024 | 914 | -55% | 0 | 0 | — |
case-12 | fail→fail | 18,712 | 5,934 | -68% | 1 | 1 | 0% | 792 | 1,047 | +32% | 0 | 0 | — |
case-13 | fail→pass | 6,367 | 1,640 | -74% | 1 | 1 | 0% | 933 | 775 | -17% | 0 | 0 | — |
case-14 | fail→pass | 12,868 | 2,138 | -83% | 1 | 1 | 0% | 770 | 861 | +12% | 0 | 0 | — |
case-15 | pass→pass | 6,941 | 1,608 | -77% | 1 | 1 | 0% | 1,117 | 751 | -33% | 0 | 0 | — |
case-16 | pass→pass | 9,328 | 8,168 | -12% | 1 | 1 | 0% | 1,548 | 1,736 | +12% | 0 | 0 | — |
case-17 | fail→pass | 8,292 | 9,428 | +14% | 1 | 1 | 0% | 1,297 | 2,025 | +56% | 0 | 0 | — |
case-18 | pass→pass | 12,323 | 11,738 | -5% | 1 | 1 | 0% | 1,658 | 2,247 | +36% | 0 | 0 | — |
case-19 | pass→pass | 16,872 | 13,552 | -20% | 1 | 1 | 0% | 2,781 | 2,791 | +0% | 0 | 0 | — |
case-20 | fail→pass | 14,243 | 6,695 | -53% | 1 | 1 | 0% | 2,133 | 1,520 | -29% | 0 | 0 | — |
case-21 | pass→pass | 10,604 | 9,642 | -9% | 1 | 1 | 0% | 1,408 | 1,804 | +28% | 0 | 0 | — |
case-22 | pass→pass | 32,561 | 30,178 | -7% | 1 | 1 | 0% | 4,733 | 4,783 | +1% | 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 19 counted toward the lift figure. The other 3 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 +32 percentage points is the difference between those two pass rates over the 19 comparable cases.
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.