What problem does it solve? Structuring a domain expert's knowledge into a rigorous, atomic knowledge graph is hard: concepts get bundled, derivation edges get confused with chronology or similarity, and premature writes corrupt the graph. This Skill turns the Agent into a disciplined interviewer and graph modeler that only writes after shared understanding is confirmed. ## Core Features & Use Cases - Dependency-ordered grilling: Builds a design tree of semantic decisions and asks the complete current frontier as numbered questions with recommended answers. - Adversarial graph review: Detects bundled concepts, hidden intermediates, cycles, parallel edges, disguised derivation, and self-referential definitions before any write. - Confirmed atomic batches: Applies changes only after explicit user authorization via derivon apply with atomic manifest replacement, crosslinking, rendering, and validation. - Use Case: A domain expert wants to convert their field knowledge into a Derivon knowledge graph; the Agent interviews them round by round, proposes candidate points and hyperedges, reviews them adversarially, and writes only confirmed batches. ## Quick Start Ask the Agent to start a Derivon creation session for your domain and answer its numbered frontier questions to confirm the first batch. ## Core Features & Use Cases
