Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →9 graph analysis operations for knowledge networks: triangles (synthesis opportunities), bridges (critical connectors), clusters (isolated subgraphs), hubs (high-degree nodes), siblings (unconnected items sharing topics), forward/backward traversal, orphans, and staleness scan. The analytical lens on your knowledge structure. Triggers on: "graph", "connections", "network analysis", "knowledge map"
.claude/skills/miosa-osa-graph/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -60% | 0% |
Analyze the knowledge graph for synthesis opportunities.
/graph # Show graph stats
/graph triangles # Find synthesis opportunities
/graph triangles --limit 10 # Limit results
/graph clusters # Find isolated components
/graph hubs # Find hub nodesbashcd engine && mix optimal.graph cd engine && mix optimal.graph triangles --limit 10 cd engine && mix optimal.graph clusters cd engine && mix optimal.graph hubs
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 9,253 | 2,131 | -77% | 1 | 1 | 0% | 1,773 | 548 | -69% | 0 | 0 | — |
case-01 | fail→pass | 9,180 | 6,359 | -31% | 1 | 1 | 0% | 1,713 | 1,219 | -29% | 0 | 0 | — |
case-02 | fail→pass | 12,229 | 7,532 | -38% | 1 | 1 | 0% | 2,080 | 1,498 | -28% | 0 | 0 | — |
case-03 | fail→pass | 7,529 | 2,235 | -70% | 1 | 1 | 0% | 1,242 | 435 | -65% | 0 | 0 | — |
case-04 | fail→pass | 11,116 | 2,069 | -81% | 1 | 1 | 0% | 2,011 | 532 | -74% | 0 | 0 | — |
case-06 | pass→pass | 14,830 | 4,830 | -67% | 1 | 1 | 0% | 2,361 | 1,016 | -57% | 0 | 0 | — |
case-07 | pass→pass | 9,414 | 1,838 | -80% | 1 | 1 | 0% | 1,617 | 476 | -71% | 0 | 0 | — |
case-08 | pass→pass | 7,342 | 1,800 | -75% | 1 | 1 | 0% | 1,251 | 501 | -60% | 0 | 0 | — |
case-09 | fail→pass | 8,255 | 2,257 | -73% | 1 | 1 | 0% | 1,506 | 596 | -60% | 0 | 0 | — |
case-10 | fail→fail | 7,783 | 3,571 | -54% | 1 | 1 | 0% | 1,378 | 697 | -49% | 0 | 0 | — |
case-11 | fail→pass | 6,577 | 1,746 | -73% | 1 | 1 | 0% | 1,162 | 479 | -59% | 0 | 0 | — |
case-12 | fail→pass | 8,099 | 1,806 | -78% | 1 | 1 | 0% | 1,424 | 497 | -65% | 0 | 0 | — |
case-13 | pass→pass | 11,454 | 3,061 | -73% | 1 | 1 | 0% | 1,768 | 720 | -59% | 0 | 0 | — |
case-14 | pass→pass | 16,641 | 9,620 | -42% | 1 | 1 | 0% | 2,689 | 1,826 | -32% | 0 | 0 | — |
case-15 | pass→pass | 13,115 | 2,340 | -82% | 1 | 1 | 0% | 2,303 | 583 | -75% | 0 | 0 | — |
case-16 | pass→pass | 4,279 | 3,228 | -25% | 1 | 1 | 0% | 736 | 766 | +4% | 0 | 0 | — |
case-17 | pass→pass | 13,119 | 9,344 | -29% | 1 | 1 | 0% | 2,124 | 1,696 | -20% | 0 | 0 | — |
case-18 | pass→pass | 12,887 | 3,270 | -75% | 1 | 1 | 0% | 2,104 | 691 | -67% | 0 | 0 | — |
case-19 | fail→pass | 18,218 | 1,495 | -92% | 1 | 1 | 0% | 1,196 | 484 | -60% | 0 | 0 | — |
case-20 | pass→pass | 11,086 | 3,286 | -70% | 1 | 1 | 0% | 1,700 | 652 | -62% | 0 | 0 | — |
case-21 | pass→pass | 5,276 | 3,686 | -30% | 1 | 1 | 0% | 895 | 692 | -23% | 0 | 0 | — |
case-22 | fail→pass | 6,307 | 2,698 | -57% | 1 | 1 | 0% | 1,117 | 543 | -51% | 0 | 0 | — |
case-23 | fail→pass | 5,332 | 4,566 | -14% | 1 | 1 | 0% | 888 | 1,000 | +13% | 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. 23 cases were attempted. The headline lift of +43 percentage points is the difference between those two pass rates over the 23 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.