Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Strategy: multi-dimensional weighted scoring and ranking — decompose a gap into independent sub-questions, then recombine into a priority list
.claude/skills/yogsoth-ai-multi-criteria-ranking/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -33% | 0% |
Multi-dimensional weighted scoring and ranking: decompose the composite question "which gap is better" into several independent dimensions, score each separately, then recombine into a final ranking via weighted summation.
Core principle: the reliability of complex judgments comes from decomposition, not holistic intuition.
Break "which gap is most worth attacking" into four independent sub-questions:
Each dimension is scored independently (1–5) to avoid cross-contamination between dimensions. Weights are set by AHP (Analytic Hierarchy Process) or specified by the user. Final score = Σ(dimension score × dimension weight).
Sensitivity check: perturb weights by ±20%; if the ranking is unchanged the conclusion is robust; if the ranking flips it must be flagged as "weight-sensitive".
| Tier | Gap count | Scoring dimensions | Sensitivity check | Final output | |------|---------|---------|-----------|---------| | S | 5–8 | ≥3 dimensions | Optional | Ranking table + attack suggestions for top 2 gaps | | M | 9–15 | ≥4 dimensions | Required | Ranking table + attack suggestions for top 3 gaps | | L | 16–20 | ≥5 dimensions | Required (multi-weight scenarios) | Ranking table + attack suggestions for top 5 gaps + weight-sensitivity report |
gap-normalization SOP: normalize input gaps into a standard format (ID, title, one-sentence description)ahp-weighting SOP: determine each dimension's weight (default: importance 0.35, feasibility 0.25, novelty 0.20, impact 0.20)importance-scoring, feasibility-scoring, novelty-scoring, impact-scoring)scoring-matrix-construction tactic: aggregate into a scoring matrixpriority-sensitivity-testing tactic: perturb weights and check ranking robustnesspriority-synthesis SOP: produce the final ranking + attack suggestionsAfter each round, record:
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | hypothesis-formation-scoring-matrix-construction | Tactic: orchestrate multi-dimensional scoring SOPs to build a comprehensive assessment matrix for all gaps | | priority-sensitivity-testing | Tactic: perturb scoring weights to test the robustness of the gap ranking against weight choice |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | gap-normalization | SOP: Unify gaps from different sources into the standard GapRecord format |
<!-- 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-04 | pass→pass | 31,129 | 31,902 | +2% | 1 | 1 | 0% | 4,164 | 5,047 | +21% | 0 | 0 | — |
case-01 | fail→pass | 50,418 | 28,382 | -44% | 1 | 1 | 0% | 8,292 | 5,585 | -33% | 0 | 0 | — |
case-02 | fail→pass | 40,880 | 39,725 | -3% | 1 | 1 | 0% | 8,256 | 7,741 | -6% | 0 | 0 | — |
case-03 | fail→fail | 43,284 | 40,218 | -7% | 1 | 1 | 0% | 8,254 | 9,092 | +10% | 0 | 0 | — |
case-05 | pass→pass | 17,517 | 48,739 | +178% | 1 | 1 | 0% | 2,708 | 3,971 | +47% | 0 | 0 | — |
case-06 | pass→pass | 24,577 | 65,892 | +168% | 1 | 1 | 0% | 4,307 | 7,338 | +70% | 0 | 0 | — |
case-07 | fail→pass | 41,994 | 27,186 | -35% | 1 | 1 | 0% | 4,296 | 5,978 | +39% | 0 | 0 | — |
case-08 | fail→pass | 33,274 | 9,007 | -73% | 1 | 1 | 0% | 904 | 1,491 | +65% | 0 | 0 | — |
case-09 | fail→fail | 17,527 | 8,126 | -54% | 1 | 1 | 0% | 1,341 | 1,364 | +2% | 0 | 0 | — |
case-10 | fail→fail | 21,300 | 11,561 | -46% | 1 | 1 | 0% | 2,392 | 2,058 | -14% | 0 | 0 | — |
case-11 | fail→pass | 20,308 | 9,338 | -54% | 1 | 1 | 0% | 2,410 | 1,609 | -33% | 0 | 0 | — |
case-12 | fail→pass | 6,249 | 3,527 | -44% | 1 | 1 | 0% | 1,019 | 1,395 | +37% | 0 | 0 | — |
case-13 | fail→pass | 16,929 | 13,807 | -18% | 1 | 1 | 0% | 2,027 | 1,902 | -6% | 0 | 0 | — |
case-14 | pass→pass | 15,910 | 15,851 | -0% | 1 | 1 | 0% | 2,376 | 2,654 | +12% | 0 | 0 | — |
case-15 | fail→fail | 13,317 | 3,519 | -74% | 1 | 1 | 0% | 1,717 | 1,520 | -11% | 0 | 0 | — |
case-16 | fail→pass | 15,469 | 4,229 | -73% | 1 | 1 | 0% | 2,437 | 1,496 | -39% | 0 | 0 | — |
case-17 | pass→pass | 10,914 | 3,826 | -65% | 1 | 1 | 0% | 1,683 | 1,569 | -7% | 0 | 0 | — |
case-18 | pass→pass | 8,631 | 5,980 | -31% | 1 | 1 | 0% | 1,310 | 1,762 | +35% | 0 | 0 | — |
case-19 | fail→pass | 14,814 | 2,315 | -84% | 1 | 1 | 0% | 2,287 | 1,261 | -45% | 0 | 0 | — |
case-20 | fail→fail | 11,796 | 3,087 | -74% | 1 | 1 | 0% | 1,855 | 1,341 | -28% | 0 | 0 | — |
case-21 | fail→pass | 34,149 | 2,426 | -93% | 1 | 1 | 0% | 2,708 | 1,264 | -53% | 0 | 0 | — |
case-22 | fail→pass | 31,248 | 35,211 | +13% | 1 | 1 | 0% | 6,419 | 7,804 | +22% | 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 20 counted toward the lift figure. The other 2 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 +50 percentage points is the difference between those two pass rates over the 20 comparable cases.
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.