What problem does it solve? When building a research wiki, ingesting papers often creates duplicate concept pages for well-known textbook material like gradient descent or transformers. This Skill pre-seeds a wiki/foundations/ directory with terminal background pages so subsequent ingestion runs deduplicate against them and create wikilinks instead of redundant concept pages. ## Core Features & Use Cases - Catalog-driven seeding: Reads a hand-curated YAML catalog of foundational concepts organized by domain (general, NLP, CV, RL) and generates one terminal page per concept. - Wikipedia-backed content: Fetches summaries and relevant sections via tools/fetch_wikipedia.py, falling back to LLM knowledge with a (LLM analysis) tag when a page is missing. - Idempotent and safe: Never overwrites existing foundation pages, writes only to wiki/foundations/, and rebuilds the wiki index and log after each run. - Use Case: Before ingesting a batch of NLP papers, run /prefill NLP to seed pages for Transformer, Attention Mechanism, and Word Embeddings so later ingestion links to them instead of duplicating them. ## Quick Start Run the prefill skill with the NLP domain to seed foundational concept pages into the wiki foundations directory.
