What problem does it solve? Turning captured source material into durable, organized knowledge is tedious and error-prone: pages get duplicated, contradictions get silently overwritten, and cross-links decay. This Skill operationalizes the ingestion of a single raw source into a structured LLM Wiki, preserving citations, links, and an audit log. ## Core Features & Use Cases - Source-to-page pipeline: Reads a file under raw/ end to end and writes a canonical source page at wiki/sources/<slug>.md with schema-compliant frontmatter. - Downstream knowledge updates: Creates or updates 5-15 entity, concept, and synthesis pages, wiring bidirectional cross-links and citations back to the source page. - Contradiction handling and audit trail: Flags conflicts with > ⚠ contradicted by callouts instead of overwriting, refreshes wiki/index.md, and appends a structured entry to wiki/log.md. - Use Case: An operation issue of type ingest names raw/karpathy-llm-wiki.md; the agent reads the space conventions, files a neutral source page, updates related concept pages, and logs the ingest so no one re-processes the same source. ## Quick Start Assign an ingest operation issue naming a file under raw/ and ask the agent to ingest it into the target wiki space.