Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use Ontoly's deterministic Software Graph and MCP capabilities for repository architecture, request tracing, dependency analysis, configuration lookup, and impact analysis before falling back to source search.
.claude/skills/mxyhi-ontoly-software-graph/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 610% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 226% | 0% |
Use this skill when a coding agent needs graph-backed understanding of a TypeScript repository. Ontoly is the source of truth; the skill only teaches the workflow.
.ontoly/, SoftwareGraph.json, diagnostics.json, validation reports, graph hash, or MCP setup.bash ontoly build .
ExplainArchitectureFindDependenciesImpactAnalysisTraceExecutionFindConfigurationUsageFrameworkReportFindDeadCodeReturn the direct answer first, then evidence:
textAuthController handles authentication. Evidence: - node: class:src/auth/auth.controller.ts:AuthController - route edges: HANDLES POST /login and POST /logout - dependency edges: USES AuthService and JwtService Confidence: high, because the graph has controller, route, and dependency edges with source locations.
NOT_FOUND with the closest graph evidence instead of inventing an answer.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→fail | 7,752 | 3,777 | -51% | 1 | 1 | 0% | 1,409 | 1,030 | -27% | 0 | 0 | — |
case-06 | fail→fail | 9,424 | 9,809 | +4% | 1 | 1 | 0% | 1,882 | 1,465 | -22% | 0 | 0 | — |
case-19 | fail→pass | 11,401 | 7,986 | -30% | 1 | 1 | 0% | 1,821 | 1,977 | +9% | 0 | 0 | — |
case-04 | fail→fail | 8,190 | 4,560 | -44% | 1 | 1 | 0% | 1,357 | 777 | -43% | 0 | 0 | — |
case-01 | fail→fail | 10,218 | 10,627 | +4% | 1 | 1 | 0% | 1,540 | 2,110 | +37% | 0 | 0 | — |
case-02 | fail→pass | 10,578 | 12,315 | +16% | 1 | 1 | 0% | 281 | 1,996 | +610% | 0 | 0 | — |
case-03 | fail→fail | 19,996 | 5,832 | -71% | 1 | 1 | 0% | 2,864 | 901 | -69% | 0 | 0 | — |
case-07 | fail→fail | 5,548 | 7,027 | +27% | 1 | 1 | 0% | 805 | 813 | +1% | 0 | 0 | — |
case-08 | fail→pass | 5,895 | 3,553 | -40% | 1 | 1 | 0% | 922 | 1,083 | +17% | 0 | 0 | — |
case-09 | fail→pass | 13,290 | 6,872 | -48% | 1 | 1 | 0% | 2,131 | 1,634 | -23% | 0 | 0 | — |
case-10 | fail→pass | 5,023 | 14,290 | +184% | 1 | 1 | 0% | 688 | 2,240 | +226% | 0 | 0 | — |
case-11 | fail→pass | 3,820 | 7,399 | +94% | 1 | 1 | 0% | 640 | 1,773 | +177% | 0 | 0 | — |
case-12 | fail→fail | 8,595 | 6,560 | -24% | 1 | 1 | 0% | 1,450 | 995 | -31% | 0 | 0 | — |
case-13 | fail→fail | 18,285 | 4,866 | -73% | 1 | 1 | 0% | 2,997 | 795 | -73% | 0 | 0 | — |
case-14 | fail→fail | 7,929 | 2,656 | -67% | 1 | 1 | 0% | 1,289 | 873 | -32% | 0 | 0 | — |
case-15 | pass→pass | 9,456 | 5,100 | -46% | 1 | 1 | 0% | 1,425 | 1,402 | -2% | 0 | 0 | — |
case-16 | pass→pass | 11,598 | 5,323 | -54% | 1 | 1 | 0% | 1,842 | 1,375 | -25% | 0 | 0 | — |
case-17 | fail→fail | 15,611 | 7,425 | -52% | 1 | 1 | 0% | 2,439 | 949 | -61% | 0 | 0 | — |
case-18 | fail→fail | 6,092 | 3,552 | -42% | 1 | 1 | 0% | 1,102 | 982 | -11% | 0 | 0 | — |
case-20 | pass→fail | 5,032 | 5,029 | -0% | 1 | 1 | 0% | 1,025 | 813 | -21% | 0 | 0 | — |
case-21 | pass→fail | 16,701 | 2,782 | -83% | 1 | 1 | 0% | 3,447 | 919 | -73% | 0 | 0 | — |
case-22 | pass→pass | 14,294 | 7,582 | -47% | 1 | 1 | 0% | 1,503 | 1,927 | +28% | 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 18 counted toward the lift figure. The other 4 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 +18 percentage points is the difference between those two pass rates over the 18 comparable cases. 3 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.