Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tactic: Trace upstream causes and downstream effects of each failure mode. Builds multi-level cause-mode-effect chains for systemic understanding.
.claude/skills/yogsoth-ai-failure-chain-tracing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-13 | ✓→✗ | ▼ Worse | -50% | 0% |
Trace each failure mode both upstream (root causes) and downstream (cascading effects) to build complete causal chains.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | failure-chain-construction | Build cause-mode-effect chains tracing upstream root causes and downstream cascading effects for each failure mode. | | severity-scoring | Rate failure mode severity 1-10 based on end-effect impact. Follows AIAG-VDA severity scale calibrated for research artifacts. |
<!-- 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-06 | pass→pass | 32,809 | 19,596 | -40% | 1 | 1 | 0% | 5,451 | 3,647 | -33% | 0 | 0 | — |
case-01 | fail→pass | 21,724 | 40,266 | +85% | 1 | 1 | 0% | 3,247 | 3,096 | -5% | 0 | 0 | — |
case-02 | fail→fail | 20,351 | 26,700 | +31% | 1 | 1 | 0% | 3,182 | 3,997 | +26% | 0 | 0 | — |
case-03 | fail→pass | 7,522 | 8,097 | +8% | 1 | 1 | 0% | 1,199 | 1,294 | +8% | 0 | 0 | — |
case-04 | fail→fail | 22,430 | 21,297 | -5% | 1 | 1 | 0% | 3,254 | 3,031 | -7% | 0 | 0 | — |
case-05 | pass→pass | 41,336 | 26,069 | -37% | 1 | 1 | 0% | 6,172 | 4,459 | -28% | 0 | 0 | — |
case-07 | pass→pass | 11,944 | 5,707 | -52% | 1 | 1 | 0% | 1,972 | 1,400 | -29% | 0 | 0 | — |
case-08 | pass→pass | 15,944 | 10,933 | -31% | 1 | 1 | 0% | 2,401 | 2,069 | -14% | 0 | 0 | — |
case-09 | fail→pass | 16,987 | 14,488 | -15% | 1 | 1 | 0% | 2,515 | 2,073 | -18% | 0 | 0 | — |
case-10 | pass→pass | 6,575 | 11,916 | +81% | 1 | 1 | 0% | 1,134 | 1,824 | +61% | 0 | 0 | — |
case-11 | pass→pass | 10,572 | 2,775 | -74% | 1 | 1 | 0% | 1,721 | 891 | -48% | 0 | 0 | — |
case-12 | pass→pass | 16,824 | 5,908 | -65% | 1 | 1 | 0% | 2,573 | 1,385 | -46% | 0 | 0 | — |
case-13 | pass→fail | 11,463 | 2,859 | -75% | 1 | 1 | 0% | 1,704 | 857 | -50% | 0 | 0 | — |
case-14 | pass→pass | 21,666 | 14,111 | -35% | 1 | 1 | 0% | 3,276 | 2,603 | -21% | 0 | 0 | — |
case-15 | pass→pass | 18,559 | 17,486 | -6% | 1 | 1 | 0% | 2,948 | 3,321 | +13% | 0 | 0 | — |
case-16 | fail→pass | 9,053 | 2,099 | -77% | 1 | 1 | 0% | 1,413 | 814 | -42% | 0 | 0 | — |
case-17 | pass→pass | 15,804 | 14,643 | -7% | 1 | 1 | 0% | 2,434 | 2,307 | -5% | 0 | 0 | — |
case-18 | pass→pass | 10,003 | 8,532 | -15% | 1 | 1 | 0% | 1,577 | 1,849 | +17% | 0 | 0 | — |
case-19 | pass→pass | 11,617 | 5,300 | -54% | 1 | 1 | 0% | 1,798 | 1,242 | -31% | 0 | 0 | — |
case-20 | pass→pass | 11,615 | 16,181 | +39% | 1 | 1 | 0% | 1,997 | 2,473 | +24% | 0 | 0 | — |
case-21 | pass→pass | 11,094 | 8,413 | -24% | 1 | 1 | 0% | 1,757 | 1,791 | +2% | 0 | 0 | — |
case-22 | pass→pass | 10,650 | 9,560 | -10% | 1 | 1 | 0% | 1,673 | 2,086 | +25% | 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 +14 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.