Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Execute one full Delphi round — collect judgments, distribute anonymous feedback, measure consensus, decide whether to continue.
.claude/skills/yogsoth-ai-iterative-convergence-round/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-03 | ✓→✗ | ▼ Worse | 0% | 0% |
Execute one complete iteration of the Delphi convergence cycle: collect independent judgments from all perspectives, distribute anonymized feedback showing the group distribution, allow revision, measure consensus level, and decide whether another round is needed.
judgment-collection to gather independent ratings/estimates from each perspectivefeedback-distribution to create anonymized summary of group responsesconsensus-measurement to compute agreement level (IQR, % agreement, Kendall's W)round-decision to determine continue/stop based on threshold and stability| SOP | Role in Tactic | |-----|---------------| | judgment-collection | Gather independent judgments from all perspectives | | feedback-distribution | Create and distribute anonymized feedback report | | consensus-measurement | Compute consensus score using appropriate method | | round-decision | Determine whether to run another round or stop |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | consensus-measurement | Compute consensus score from collected judgments using the appropriate statistical method. | | feedback-distribution | Create anonymized feedback report summarizing group judgment distribution for a given round. | | judgment-collection | Collect independent judgments from all perspectives on a given question. | | round-decision | Decide whether to continue iterating or stop based on consensus score, round number, and stability. |
<!-- 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 | 19,985 | 70,278 | +252% | 1 | 1 | 0% | 2,346 | 2,053 | -12% | 0 | 0 | — |
case-02 | fail→pass | 38,492 | 64,879 | +69% | 1 | 1 | 0% | 2,090 | 2,086 | -0% | 0 | 0 | — |
case-03 | pass→fail | 19,105 | 17,846 | -7% | 1 | 1 | 0% | 2,199 | 2,198 | -0% | 0 | 0 | — |
case-04 | pass→pass | 16,350 | 12,384 | -24% | 1 | 1 | 0% | 1,680 | 1,535 | -9% | 0 | 0 | — |
case-05 | pass→pass | 19,045 | 18,059 | -5% | 1 | 1 | 0% | 1,099 | 2,091 | +90% | 0 | 0 | — |
case-06 | pass→fail | 17,018 | 14,335 | -16% | 1 | 1 | 0% | 2,002 | 2,140 | +7% | 0 | 0 | — |
case-07 | fail→pass | 13,459 | 27,549 | +105% | 1 | 1 | 0% | 2,136 | 1,943 | -9% | 0 | 0 | — |
case-08 | pass→pass | 19,221 | 17,079 | -11% | 1 | 1 | 0% | 2,018 | 1,756 | -13% | 0 | 0 | — |
case-09 | fail→fail | 18,650 | 2,965 | -84% | 1 | 1 | 0% | 1,455 | 988 | -32% | 0 | 0 | — |
case-10 | pass→pass | 22,214 | 18,186 | -18% | 1 | 1 | 0% | 2,087 | 2,134 | +2% | 0 | 0 | — |
case-11 | pass→pass | 12,218 | 3,058 | -75% | 1 | 1 | 0% | 694 | 1,116 | +61% | 0 | 0 | — |
case-12 | pass→pass | 21,635 | 14,905 | -31% | 1 | 1 | 0% | 2,134 | 2,151 | +1% | 0 | 0 | — |
case-13 | pass→pass | 15,069 | 6,836 | -55% | 1 | 1 | 0% | 2,230 | 1,572 | -30% | 0 | 0 | — |
case-14 | pass→pass | 16,322 | 10,846 | -34% | 1 | 1 | 0% | 1,855 | 1,958 | +6% | 0 | 0 | — |
case-15 | pass→pass | 14,090 | 11,114 | -21% | 1 | 1 | 0% | 1,140 | 1,459 | +28% | 0 | 0 | — |
case-16 | pass→pass | 27,427 | 15,605 | -43% | 1 | 1 | 0% | 2,034 | 1,138 | -44% | 0 | 0 | — |
case-17 | pass→pass | 5,317 | 9,511 | +79% | 1 | 1 | 0% | 838 | 818 | -2% | 0 | 0 | — |
case-18 | pass→pass | 7,170 | 4,426 | -38% | 1 | 1 | 0% | 1,122 | 900 | -20% | 0 | 0 | — |
case-19 | pass→pass | 10,128 | 3,417 | -66% | 1 | 1 | 0% | 1,713 | 791 | -54% | 0 | 0 | — |
case-25 | pass→fail | 11,137 | 26,688 | +140% | 1 | 1 | 0% | 2,576 | 6,864 | +166% | 0 | 0 | — |
case-20 | fail→pass | 14,937 | 10,461 | -30% | 1 | 1 | 0% | 2,198 | 1,886 | -14% | 0 | 0 | — |
case-21 | pass→pass | 14,492 | 9,570 | -34% | 1 | 1 | 0% | 1,382 | 868 | -37% | 0 | 0 | — |
case-22 | pass→pass | 31,173 | 6,833 | -78% | 1 | 1 | 0% | 991 | 1,351 | +36% | 0 | 0 | — |
case-23 | pass→pass | 15,451 | 10,845 | -30% | 1 | 1 | 0% | 1,735 | 2,227 | +28% | 0 | 0 | — |
case-24 | pass→pass | 15,964 | 69,356 | +334% | 1 | 1 | 0% | 2,704 | 6,428 | +138% | 0 | 0 | — |
case-26 | pass→fail | 14,243 | 17,882 | +26% | 1 | 1 | 0% | 2,579 | 3,955 | +53% | 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. 26 cases were attempted. The headline lift of 0 percentage points is the difference between those two pass rates over the 26 comparable cases. 4 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.