Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Collect independent judgments from all perspectives on a given question.
.claude/skills/yogsoth-ai-judgment-collection/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 80% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 717% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 39% | 0% |
Collect independent judgments from each perspective on the focal question. Each perspective provides its rating/estimate and supporting reasoning without seeing others' responses.
Spawn a subagent that takes the question and list of perspectives, then generates an independent judgment from each perspective's viewpoint. Judgments include both a structured response (rating, probability, or position) and free-text reasoning.
Output MUST contain: one judgment object per perspective, each with perspective_id, response, and reasoning fields. Missing any perspective fails the gate.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. Used by SOPs that declare execution: subagent. |
<!-- 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-17 | fail→pass | 16,227 | 15,621 | -4% | 1 | 1 | 0% | 2,728 | 2,240 | -18% | 0 | 0 | — |
case-05 | pass→fail | 29,481 | 29,049 | -1% | 1 | 1 | 0% | 2,129 | 5,420 | +155% | 0 | 0 | — |
case-01 | fail→pass | 49,760 | 83,059 | +67% | 1 | 1 | 0% | 1,539 | 2,764 | +80% | 0 | 0 | — |
case-02 | fail→pass | 50,020 | 76,490 | +53% | 1 | 1 | 0% | 1,988 | 2,703 | +36% | 0 | 0 | — |
case-03 | pass→pass | 23,721 | 25,122 | +6% | 1 | 1 | 0% | 2,138 | 2,995 | +40% | 0 | 0 | — |
case-04 | pass→pass | 22,254 | 37,333 | +68% | 1 | 1 | 0% | 2,732 | 4,477 | +64% | 0 | 0 | — |
case-06 | fail→pass | 5,529 | 96,570 | +1647% | 1 | 1 | 0% | 945 | 7,720 | +717% | 0 | 0 | — |
case-07 | fail→fail | 19,754 | 32,540 | +65% | 1 | 1 | 0% | 2,031 | 907 | -55% | 0 | 0 | — |
case-08 | fail→pass | 12,992 | 24,206 | +86% | 1 | 1 | 0% | 1,933 | 2,694 | +39% | 0 | 0 | — |
case-09 | fail→pass | 12,295 | 70,882 | +477% | 1 | 1 | 0% | 1,162 | 3,711 | +219% | 0 | 0 | — |
case-10 | fail→fail | 14,564 | 22,801 | +57% | 1 | 1 | 0% | 1,525 | 1,394 | -9% | 0 | 0 | — |
case-11 | fail→fail | 22,197 | 30,583 | +38% | 1 | 1 | 0% | 2,512 | 765 | -70% | 0 | 0 | — |
case-12 | fail→pass | 13,449 | 15,907 | +18% | 1 | 1 | 0% | 1,506 | 2,692 | +79% | 0 | 0 | — |
case-13 | fail→pass | 13,930 | 39,112 | +181% | 1 | 1 | 0% | 2,309 | 2,528 | +9% | 0 | 0 | — |
case-14 | fail→pass | 11,318 | 28,071 | +148% | 1 | 1 | 0% | 1,783 | 2,835 | +59% | 0 | 0 | — |
case-15 | fail→pass | 18,933 | 47,823 | +153% | 1 | 1 | 0% | 2,760 | 2,875 | +4% | 0 | 0 | — |
case-16 | fail→pass | 13,543 | 18,427 | +36% | 1 | 1 | 0% | 1,469 | 3,066 | +109% | 0 | 0 | — |
case-18 | fail→fail | 18,997 | 20,058 | +6% | 1 | 1 | 0% | 3,073 | 838 | -73% | 0 | 0 | — |
case-19 | fail→pass | 13,993 | 13,337 | -5% | 1 | 1 | 0% | 2,051 | 2,151 | +5% | 0 | 0 | — |
case-20 | fail→pass | 17,338 | 18,230 | +5% | 1 | 1 | 0% | 2,535 | 2,944 | +16% | 0 | 0 | — |
case-21 | pass→pass | 15,080 | 24,784 | +64% | 1 | 1 | 0% | 2,205 | 2,944 | +34% | 0 | 0 | — |
case-22 | fail→fail | 13,573 | 13,054 | -4% | 1 | 1 | 0% | 2,029 | 836 | -59% | 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 17 counted toward the lift figure. The other 5 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 +55 percentage points is the difference between those two pass rates over the 17 comparable cases. 6 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.