What problem does it solve? Karpathy-pattern LLM wikis accumulate dozens of markdown articles, raw sources, and wikilinks that are hard to navigate as a whole. This Skill turns such a wiki into a structured knowledge graph with entities, claims, topics, and implicit relationships, so you can explore the knowledge base visually instead of reading files one by one. ## Core Features & Use Cases - Deterministic Wiki Parsing: Detects the Karpathy three-layer pattern (raw sources, wiki markdown, schema file), extracts wikilinks, headings, frontmatter, and categories from index.md into a scan manifest. - LLM-Based Implicit Analysis: Dispatches analyzer subagents in batches to extract entities, claims, and implicit cross-references that wikilinks alone do not capture. - Graph Assembly and Dashboard: Merges scan and analysis results with entity deduplication and edge normalization, builds layers and a guided tour from index.md categories, then auto-triggers the dashboard. - Use Case: Point it at a research wiki with 50 markdown articles and a raw/ folder of papers; it produces a knowledge-graph.json with articles, entities, topics, and claims ready for force-directed visualization. ## Quick Start Run the understand-knowledge skill on my wiki directory to build an interactive knowledge graph of all articles and their relationships.
