Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Systematically sweep consensus thresholds to observe which items achieve consensus at what level, producing a threshold-consensus curve.
.claude/skills/yogsoth-ai-threshold-calibration/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -52% | 0% |
Systematically vary the consensus threshold to understand the sensitivity of consensus classification. Rather than picking a single arbitrary threshold, sweep across a range to see which items are robust consensus (agree at any threshold) vs. fragile (only consensus at lenient thresholds).
threshold-sweep to compute consensus status at multiple threshold levelsconsensus-classification to categorize items at the chosen operating thresholdconsensus-measurement to validate final consensus scores| SOP | Role in Tactic | |-----|---------------| | threshold-sweep | Compute consensus at multiple threshold levels, produce curve | | consensus-classification | Classify items as consensus/dissensus at operating threshold | | consensus-measurement | Validate final consensus scores with appropriate method |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | consensus-classification | Classify items as consensus or dissensus at a given threshold. | | consensus-measurement | Compute consensus score from collected judgments using the appropriate statistical method. | | threshold-sweep | Compute consensus status at multiple threshold levels to produce a threshold-consensus curve. |
<!-- 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 | 22,659 | 13,478 | -41% | 1 | 1 | 0% | 2,875 | 1,865 | -35% | 0 | 0 | — |
case-02 | pass→pass | 14,915 | 11,002 | -26% | 1 | 1 | 0% | 1,784 | 1,267 | -29% | 0 | 0 | — |
case-03 | fail→pass | 23,188 | 18,502 | -20% | 1 | 1 | 0% | 3,028 | 2,084 | -31% | 0 | 0 | — |
case-04 | pass→pass | 14,792 | 14,184 | -4% | 1 | 1 | 0% | 1,582 | 1,244 | -21% | 0 | 0 | — |
case-05 | pass→pass | 19,750 | 16,101 | -18% | 1 | 1 | 0% | 2,225 | 1,952 | -12% | 0 | 0 | — |
case-06 | fail→pass | 20,126 | 17,002 | -16% | 1 | 1 | 0% | 2,300 | 2,597 | +13% | 0 | 0 | — |
case-07 | pass→pass | 27,740 | 2,826 | -90% | 1 | 1 | 0% | 2,312 | 845 | -63% | 0 | 0 | — |
case-08 | fail→pass | 38,634 | 8,131 | -79% | 1 | 1 | 0% | 1,306 | 923 | -29% | 0 | 0 | — |
case-09 | pass→pass | 15,791 | 2,629 | -83% | 1 | 1 | 0% | 1,877 | 849 | -55% | 0 | 0 | — |
case-10 | fail→pass | 19,207 | 8,925 | -54% | 1 | 1 | 0% | 2,328 | 1,112 | -52% | 0 | 0 | — |
case-11 | fail→pass | 11,376 | 3,738 | -67% | 1 | 1 | 0% | 1,944 | 1,081 | -44% | 0 | 0 | — |
case-12 | fail→pass | 17,807 | 12,770 | -28% | 1 | 1 | 0% | 2,015 | 1,650 | -18% | 0 | 0 | — |
case-13 | pass→pass | 20,300 | 14,486 | -29% | 1 | 1 | 0% | 2,352 | 1,985 | -16% | 0 | 0 | — |
case-14 | fail→pass | 27,062 | 14,356 | -47% | 1 | 1 | 0% | 3,760 | 2,119 | -44% | 0 | 0 | — |
case-15 | pass→pass | 17,749 | 11,862 | -33% | 1 | 1 | 0% | 2,128 | 1,422 | -33% | 0 | 0 | — |
case-16 | fail→pass | 16,884 | 9,252 | -45% | 1 | 1 | 0% | 1,636 | 1,107 | -32% | 0 | 0 | — |
case-17 | fail→pass | 18,645 | 9,486 | -49% | 1 | 1 | 0% | 2,377 | 1,210 | -49% | 0 | 0 | — |
case-18 | fail→fail | 20,474 | 12,275 | -40% | 1 | 1 | 0% | 2,455 | 2,429 | -1% | 0 | 0 | — |
case-19 | pass→fail | 15,636 | 15,413 | -1% | 1 | 1 | 0% | 3,336 | 2,525 | -24% | 0 | 0 | — |
case-20 | pass→fail | 10,714 | 19,971 | +86% | 1 | 1 | 0% | 1,887 | 2,905 | +54% | 0 | 0 | — |
case-21 | pass→pass | 17,244 | 19,518 | +13% | 1 | 1 | 0% | 3,132 | 3,232 | +3% | 0 | 0 | — |
case-22 | pass→pass | 17,180 | 12,267 | -29% | 1 | 1 | 0% | 1,897 | 1,628 | -14% | 0 | 0 | — |
case-23 | pass→pass | 14,271 | 10,862 | -24% | 1 | 1 | 0% | 2,279 | 2,096 | -8% | 0 | 0 | — |
case-24 | pass→pass | 10,743 | 11,651 | +8% | 1 | 1 | 0% | 1,564 | 1,465 | -6% | 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. 24 cases were attempted, and 23 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 +33 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 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.