Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Weaviate — open-source vector database with built-in ML. Hybrid search (vector + keyword), generative search, graph connections, multi-modal (text + image), and automatic schema inference.
.claude/skills/mkurman-weaviate/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -41% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 13% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 24% | 0% |
Weaviate is an open-source vector database with built-in vectorization modules (OpenAI, Cohere, HuggingFace, Transformers, CLIP, multi-modal). Supports hybrid search (vector + BM25 keyword), generative search (RAG with LLM integration), and multi-modal data.
bashdocker run -p 8080:8080 semitechnologies/weaviate:latest
pythonimport weaviate import weaviate.classes as wvc client = weaviate.connect_to_local() collection = client.collections.create( name="Documents", vectorizer_config=wvc.config.Configure.Vectorizer.text2vec_transformers(), ) collection.data.insert({ "title": "Paris", "content": "Paris is the capital of France. It is known for the Eiffel Tower.", }) # Hybrid search (vector + keyword) response = collection.query.hybrid(query="French capital", limit=5) for obj in response.objects: print(obj.properties)
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 9,034 | 5,324 | -41% | 1 | 1 | 0% | 1,846 | 1,359 | -26% | 0 | 0 | — |
case-02 | fail→pass | 5,217 | 3,348 | -36% | 1 | 1 | 0% | 934 | 894 | -4% | 0 | 0 | — |
case-03 | fail→fail | 5,673 | 4,481 | -21% | 1 | 1 | 0% | 1,023 | 1,134 | +11% | 0 | 0 | — |
case-04 | pass→pass | 8,791 | 3,490 | -60% | 1 | 1 | 0% | 1,509 | 894 | -41% | 0 | 0 | — |
case-05 | pass→pass | 2,981 | 2,047 | -31% | 1 | 1 | 0% | 528 | 596 | +13% | 0 | 0 | — |
case-06 | pass→pass | 4,570 | 3,670 | -20% | 1 | 1 | 0% | 762 | 943 | +24% | 0 | 0 | — |
case-07 | pass→pass | 4,993 | 1,470 | -71% | 1 | 1 | 0% | 915 | 529 | -42% | 0 | 0 | — |
case-08 | pass→pass | 3,836 | 2,407 | -37% | 1 | 1 | 0% | 681 | 705 | +4% | 0 | 0 | — |
case-09 | pass→pass | 4,253 | 1,689 | -60% | 1 | 1 | 0% | 744 | 551 | -26% | 0 | 0 | — |
case-10 | pass→pass | 4,269 | 2,602 | -39% | 1 | 1 | 0% | 794 | 733 | -8% | 0 | 0 | — |
case-11 | pass→pass | 2,162 | 2,048 | -5% | 1 | 1 | 0% | 412 | 628 | +52% | 0 | 0 | — |
case-12 | pass→pass | 5,170 | 2,140 | -59% | 1 | 1 | 0% | 943 | 641 | -32% | 0 | 0 | — |
case-13 | pass→pass | 4,908 | 4,339 | -12% | 1 | 1 | 0% | 864 | 1,014 | +17% | 0 | 0 | — |
case-14 | pass→pass | 4,253 | 3,422 | -20% | 1 | 1 | 0% | 634 | 933 | +47% | 0 | 0 | — |
case-15 | pass→pass | 3,797 | 3,223 | -15% | 1 | 1 | 0% | 570 | 793 | +39% | 0 | 0 | — |
case-16 | pass→pass | 12,093 | 8,306 | -31% | 1 | 1 | 0% | 2,275 | 1,932 | -15% | 0 | 0 | — |
case-17 | pass→pass | 2,788 | 2,431 | -13% | 1 | 1 | 0% | 424 | 645 | +52% | 0 | 0 | — |
case-18 | pass→pass | 4,364 | 4,142 | -5% | 1 | 1 | 0% | 726 | 973 | +34% | 0 | 0 | — |
case-19 | pass→pass | 3,025 | 1,798 | -41% | 1 | 1 | 0% | 434 | 549 | +26% | 0 | 0 | — |
case-20 | pass→pass | 12,068 | 8,681 | -28% | 1 | 1 | 0% | 2,521 | 2,257 | -10% | 0 | 0 | — |
case-21 | pass→pass | 8,609 | 4,377 | -49% | 1 | 1 | 0% | 1,543 | 1,119 | -27% | 0 | 0 | — |
case-22 | pass→pass | 13,238 | 7,770 | -41% | 1 | 1 | 0% | 2,746 | 1,810 | -34% | 0 | 0 | — |
case-23 | pass→pass | 6,897 | 4,232 | -39% | 1 | 1 | 0% | 1,275 | 1,100 | -14% | 0 | 0 | — |
case-24 | pass→pass | 12,253 | 7,752 | -37% | 1 | 1 | 0% | 2,208 | 1,692 | -23% | 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. 24 cases were attempted. The headline lift of +8 percentage points is the difference between those two pass rates over the 24 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.