Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Aggregate probability judgments across perspectives using Real-Time Delphi or prediction market mechanisms.
.claude/skills/yogsoth-ai-futures-calibration/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 144% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -28% | 0% |
Purpose: Aggregate probabilistic forecasts from multiple perspectives into calibrated predictions. Uses Real-Time Delphi continuous updating or prediction market scoring to produce well-calibrated probability estimates for future events.
When to use:
| Parameter | Constraint | |-----------|-----------| | Rounds | 2–3 (continuous updating preferred) | | Perspectives | ≥4 independent forecasters | | Calibration target | Brier score improvement across rounds |
| Key | Type | Description | |-----|------|-------------| | question | string | The forecasting question | | time_horizon | string | When the event would resolve | | perspectives | array | List of forecaster perspectives | | rounds | array | History of probability estimates | | current_aggregate | float | Current aggregated probability | | calibration_metrics | object | Brier score, log score tracking |
yamlforecast_question: <question> time_horizon: <date/period> calibrated_probability: <float 0-1> confidence_interval: [lower, upper] rounds_completed: <int> convergence_pattern: converging/stable/polarizing key_considerations: [...]
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | iterative-convergence-round | Execute one full Delphi round — collect judgments, distribute anonymous feedback, measure consensus, decide whether to continue. | | threshold-calibration | Systematically sweep consensus thresholds to observe which items achieve consensus at what level, producing a threshold-consensus curve. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | consensus-synthesis | Synthesize all rounds into a final consensus report documenting agreements, dissent, and process. |
<!-- 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 | 12,100 | 26,754 | +121% | 1 | 1 | 0% | 2,131 | 3,397 | +59% | 0 | 0 | — |
case-02 | fail→pass | 8,297 | 10,817 | +30% | 1 | 1 | 0% | 1,575 | 2,604 | +65% | 0 | 0 | — |
case-03 | fail→pass | 13,164 | 16,254 | +23% | 1 | 1 | 0% | 2,169 | 3,392 | +56% | 0 | 0 | — |
case-04 | pass→pass | 15,356 | 7,766 | -49% | 1 | 1 | 0% | 2,274 | 1,786 | -21% | 0 | 0 | — |
case-05 | pass→pass | 10,442 | 3,314 | -68% | 1 | 1 | 0% | 1,654 | 1,135 | -31% | 0 | 0 | — |
case-06 | pass→pass | 14,132 | 4,464 | -68% | 1 | 1 | 0% | 2,277 | 1,302 | -43% | 0 | 0 | — |
case-07 | pass→pass | 12,860 | 6,175 | -52% | 1 | 1 | 0% | 1,796 | 1,525 | -15% | 0 | 0 | — |
case-08 | fail→pass | 30,071 | 11,610 | -61% | 1 | 1 | 0% | 1,164 | 2,841 | +144% | 0 | 0 | — |
case-09 | fail→pass | 8,505 | 2,188 | -74% | 1 | 1 | 0% | 1,340 | 967 | -28% | 0 | 0 | — |
case-10 | fail→pass | 9,552 | 4,643 | -51% | 1 | 1 | 0% | 1,737 | 1,521 | -12% | 0 | 0 | — |
case-11 | pass→pass | 13,013 | 10,520 | -19% | 1 | 1 | 0% | 2,133 | 2,293 | +8% | 0 | 0 | — |
case-12 | fail→pass | 7,002 | 1,480 | -79% | 1 | 1 | 0% | 1,118 | 763 | -32% | 0 | 0 | — |
case-13 | fail→pass | 9,507 | 1,927 | -80% | 1 | 1 | 0% | 1,470 | 880 | -40% | 0 | 0 | — |
case-14 | pass→pass | 9,913 | 1,674 | -83% | 1 | 1 | 0% | 1,450 | 881 | -39% | 0 | 0 | — |
case-15 | pass→pass | 8,321 | 1,797 | -78% | 1 | 1 | 0% | 1,328 | 804 | -39% | 0 | 0 | — |
case-16 | fail→pass | 11,029 | 3,071 | -72% | 1 | 1 | 0% | 1,789 | 1,135 | -37% | 0 | 0 | — |
case-17 | fail→pass | 11,651 | 4,429 | -62% | 1 | 1 | 0% | 2,210 | 1,374 | -38% | 0 | 0 | — |
case-18 | pass→pass | 3,722 | 2,074 | -44% | 1 | 1 | 0% | 519 | 866 | +67% | 0 | 0 | — |
case-19 | pass→pass | 14,558 | 7,871 | -46% | 1 | 1 | 0% | 2,260 | 1,845 | -18% | 0 | 0 | — |
case-20 | pass→fail | 13,644 | 12,904 | -5% | 1 | 1 | 0% | 2,692 | 2,851 | +6% | 0 | 0 | — |
case-21 | pass→pass | 12,951 | 11,472 | -11% | 1 | 1 | 0% | 2,480 | 2,823 | +14% | 0 | 0 | — |
case-22 | pass→pass | 15,191 | 15,335 | +1% | 1 | 1 | 0% | 2,526 | 3,276 | +30% | 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 21 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 +41 percentage points is the difference between those two pass rates over the 21 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.