Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Compute consensus score from collected judgments using the appropriate statistical method.
.claude/skills/yogsoth-ai-consensus-measurement/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 202% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 44% | 0% |
Compute a quantitative consensus score from the collected judgments. Automatically selects the appropriate measurement method based on data type (IQR for continuous, percentage agreement for categorical, Kendall's W for rankings).
Spawn a subagent that analyzes the judgments array, determines the appropriate consensus metric, computes the score, and reports whether the consensus threshold is met.
Output MUST contain: consensus_score (numeric), method_used (string), threshold_met (boolean), and interpretation (string). Score must be computed, not estimated.
<!-- 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-01 | fail→fail | 10,254 | 12,956 | +26% | 1 | 1 | 0% | 1,530 | 2,492 | +63% | 0 | 0 | — |
case-02 | fail→pass | 7,823 | 25,943 | +232% | 1 | 1 | 0% | 1,183 | 3,577 | +202% | 0 | 0 | — |
case-03 | fail→pass | 15,703 | 24,797 | +58% | 1 | 1 | 0% | 2,913 | 4,807 | +65% | 0 | 0 | — |
case-04 | fail→fail | 13,706 | 22,446 | +64% | 1 | 1 | 0% | 2,225 | 3,287 | +48% | 0 | 0 | — |
case-05 | pass→pass | 9,042 | 14,517 | +61% | 1 | 1 | 0% | 1,344 | 1,865 | +39% | 0 | 0 | — |
case-06 | fail→pass | 11,895 | 15,750 | +32% | 1 | 1 | 0% | 2,307 | 3,338 | +45% | 0 | 0 | — |
case-07 | fail→pass | 17,987 | 11,663 | -35% | 1 | 1 | 0% | 3,481 | 2,433 | -30% | 0 | 0 | — |
case-08 | fail→pass | 8,924 | 10,172 | +14% | 1 | 1 | 0% | 1,421 | 2,040 | +44% | 0 | 0 | — |
case-09 | fail→fail | 12,510 | 18,481 | +48% | 1 | 1 | 0% | 2,269 | 3,734 | +65% | 0 | 0 | — |
case-10 | fail→pass | 7,539 | 11,813 | +57% | 1 | 1 | 0% | 1,361 | 2,428 | +78% | 0 | 0 | — |
case-11 | fail→pass | 7,327 | 9,025 | +23% | 1 | 1 | 0% | 1,191 | 1,806 | +52% | 0 | 0 | — |
case-12 | fail→pass | 10,950 | 21,691 | +98% | 1 | 1 | 0% | 2,128 | 3,243 | +52% | 0 | 0 | — |
case-13 | fail→fail | 23,241 | 21,042 | -9% | 1 | 1 | 0% | 4,538 | 4,200 | -7% | 0 | 0 | — |
case-14 | pass→pass | 7,343 | 7,272 | -1% | 1 | 1 | 0% | 1,135 | 1,515 | +33% | 0 | 0 | — |
case-15 | pass→pass | 9,903 | 21,625 | +118% | 1 | 1 | 0% | 2,106 | 4,753 | +126% | 0 | 0 | — |
case-16 | fail→pass | 11,604 | 15,832 | +36% | 1 | 1 | 0% | 2,338 | 3,343 | +43% | 0 | 0 | — |
case-17 | pass→fail | 7,052 | 13,098 | +86% | 1 | 1 | 0% | 1,198 | 1,032 | -14% | 0 | 0 | — |
case-18 | fail→fail | 12,171 | 15,630 | +28% | 1 | 1 | 0% | 2,231 | 3,146 | +41% | 0 | 0 | — |
case-19 | fail→pass | 11,927 | 12,289 | +3% | 1 | 1 | 0% | 2,248 | 2,428 | +8% | 0 | 0 | — |
case-20 | pass→pass | 2,300 | 10,093 | +339% | 1 | 1 | 0% | 443 | 2,027 | +358% | 0 | 0 | — |
case-21 | pass→pass | 12,455 | 6,055 | -51% | 1 | 1 | 0% | 2,635 | 1,270 | -52% | 0 | 0 | — |
case-22 | pass→pass | 17,608 | 16,045 | -9% | 1 | 1 | 0% | 2,485 | 2,745 | +10% | 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.