What problem does it solve? Designing a structural knowledge graph over a real codebase often produces node and edge types based on assumptions rather than verified code, leading to ontologies that drift from the actual architecture. This Skill guides an interactive, evidence-grounded design process for Neo4j-style knowledge graphs and their MCP tool layers, with a mandatory dry-run validation against real code before any ontology file is written. ## Core Features & Use Cases - Demand-driven ontology design: Collects the 5-10 real questions the graph must answer before proposing any node types, relationship triples, or properties. - Mandatory dry-run validation: Manually checks 2-3 real cascades in the actual codebase against declared triples and reports the outcome whether problems are found or the pass is clean. - MCP tool tiering: Maps each target question to a Tier-1 curated tool with typed parameters or an explicit Tier-0 read_cypher fallback, including what an empty result means for each tool. - Use Case: When extending the ECommerceApp knowledge graph (modules, endpoints, entities, Outbox/Inbox message flows), use this Skill to propose only the delta to the confirmed ontology, verify it against the code, and then update tools/kg/seed/ontology.json and .cypher files. ## Quick Start Ask the assistant to design a knowledge-graph ontology for this codebase that answers which modules break when a given message contract changes.
