This orchestration layer sits between Claude and the memory MCP server to handle the messy parts of knowledge graph queries: parsing natural language into entity searches, deduplicating results above 85% similarity, and boosting scores when entities cross-reference each other. It ranks everything by recency times relevance times source authority, then hands back unified results. You'd reach for this when basic memory search isn't cutting it and you need multi-hop graph traversal or entity relationship tracking. The complexity is high and it's not user-invocable, so it's really infrastructure for other skills that need smarter memory retrieval. Think of it as a query optimization layer that prevents you from missing relevant context scattered across your knowledge graph.
npx skills add https://github.com/yonatangross/orchestkit --skill memory-fabric