Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tactic for generating and populating combination matrices — cross dimensions to enumerate the design space.
.claude/skills/yogsoth-ai-matrix-generation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 89% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 318% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 252% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -22% | 0% |
Generate combination matrices by crossing dimensions. Populate cells with existing work, mark empty cells as opportunities, flag impossible combinations.
<HARD-GATE> ≥1 matrix generated with ≥50% of cells classified (occupied/empty/impossible) per invocation. </HARD-GATE>
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | combination-enumeration | SOP for systematically enumerating combinations across dimensions. | | knowledge-structuring-novelty-scoring | SOP for scoring empty cells by novelty potential — how surprising and valuable would this combination be? | | matrix-export | SOP for exporting the dimensional matrix as a readable document or structured data. | | question-generation | SOP for generating research questions from promising gaps in the design space. |
<!-- 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 | 18,646 | 18,718 | +0% | 1 | 1 | 0% | 1,986 | 2,538 | +28% | 0 | 0 | — |
case-01 | fail→pass | 29,263 | 48,349 | +65% | 1 | 1 | 0% | 4,536 | 8,584 | +89% | 0 | 0 | — |
case-02 | fail→pass | 8,993 | 41,663 | +363% | 1 | 1 | 0% | 1,854 | 7,756 | +318% | 0 | 0 | — |
case-03 | fail→fail | 22,577 | 18,394 | -19% | 1 | 1 | 0% | 2,007 | 3,393 | +69% | 0 | 0 | — |
case-05 | pass→pass | 20,030 | 19,053 | -5% | 1 | 1 | 0% | 2,288 | 2,470 | +8% | 0 | 0 | — |
case-06 | pass→pass | 11,175 | 11,873 | +6% | 1 | 1 | 0% | 1,985 | 1,650 | -17% | 0 | 0 | — |
case-07 | pass→pass | 23,187 | 25,630 | +11% | 1 | 1 | 0% | 3,000 | 3,448 | +15% | 0 | 0 | — |
case-08 | pass→pass | 23,971 | 21,420 | -11% | 1 | 1 | 0% | 2,854 | 2,976 | +4% | 0 | 0 | — |
case-09 | fail→pass | 64,182 | 19,786 | -69% | 1 | 1 | 0% | 1,054 | 3,705 | +252% | 0 | 0 | — |
case-10 | fail→fail | 23,158 | 46,478 | +101% | 1 | 1 | 0% | 3,461 | 8,576 | +148% | 0 | 0 | — |
case-11 | pass→pass | 14,468 | 21,630 | +50% | 1 | 1 | 0% | 1,481 | 3,013 | +103% | 0 | 0 | — |
case-12 | pass→pass | 16,731 | 11,138 | -33% | 1 | 1 | 0% | 1,799 | 2,254 | +25% | 0 | 0 | — |
case-13 | pass→pass | 21,305 | 25,018 | +17% | 1 | 1 | 0% | 2,776 | 3,762 | +36% | 0 | 0 | — |
case-14 | pass→pass | 19,015 | 15,276 | -20% | 1 | 1 | 0% | 2,032 | 2,049 | +1% | 0 | 0 | — |
case-15 | pass→pass | 23,554 | 28,813 | +22% | 1 | 1 | 0% | 3,133 | 4,207 | +34% | 0 | 0 | — |
case-16 | fail→pass | 15,726 | 37,434 | +138% | 1 | 1 | 0% | 2,226 | 3,099 | +39% | 0 | 0 | — |
case-17 | fail→pass | 13,439 | 4,401 | -67% | 1 | 1 | 0% | 1,411 | 1,103 | -22% | 0 | 0 | — |
case-18 | pass→pass | 27,814 | 30,918 | +11% | 1 | 1 | 0% | 3,558 | 4,340 | +22% | 0 | 0 | — |
case-19 | fail→pass | 20,863 | 16,013 | -23% | 1 | 1 | 0% | 2,458 | 2,880 | +17% | 0 | 0 | — |
case-20 | pass→pass | 21,955 | 32,103 | +46% | 1 | 1 | 0% | 1,910 | 6,047 | +217% | 0 | 0 | — |
case-21 | pass→pass | 14,029 | 18,578 | +32% | 1 | 1 | 0% | 2,485 | 2,760 | +11% | 0 | 0 | — |
case-22 | pass→pass | 14,828 | 17,325 | +17% | 1 | 1 | 0% | 2,024 | 2,926 | +45% | 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 +27 percentage points is the difference between those two pass rates over the 21 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.