What problem does it solve? It answers user questions strictly from an accumulated wiki knowledge base instead of re-reading raw sources or relying on ungrounded model knowledge, ensuring every claim is cited and gaps are honestly surfaced. ## Core Features & Use Cases - Graph-Based Retrieval: Traverses wiki entities and edges via read-only CLI subcommands (list-entities, read-meta, read-edges) to build a relevant subgraph for each question. - Cited Synthesis: Produces answers with direct answers, supporting evidence cited as wiki page links, confidence calibration, and explicit gap statements. - Optional Output Filing: With user confirmation, files answers as wiki output or summary pages with produced edges and log entries. - Use Case: A user asks "What do the wiki sources say about softmax temperature scaling?" The skill locates the concept page, follows inbound edges to all citing sources, and returns a synthesized, cited answer without touching raw files. ## Quick Start Ask a question about any topic already ingested into the wiki, such as what the wiki says about a specific concept or a comparison across recent sources.
