Skill

Create a compounding knowledge base—a workspace wiki of plain markdown pages linked together with wikilinks, designed to grow denser and more interconnected as you add sources. Ingest articles, documents, and notes into versioned raw files, then synthesize them into organized pages with frontmatter, cross-references, and provenance tracking. Run automated linting to catch broken links, orphaned pages, and schema drift, keeping your knowledge graph clean and queryable.

llm-wiki creates and updates a structured markdown documentation system for any codebase, storing it in a `.llm-wiki/` directory at your project root. It handles reconnaissance, skeleton generation, article writing, cross-linking, and linting—letting humans direct while the LLM manages the bookkeeping. The wiki compounds over time as your codebase evolves, capturing architecture, modules, concepts, APIs, and guides in an interlinked format.

wiki-retrieve upgrades vault search from page-level to chunk-level granularity, combining contextual prefixes, BM25 sparse retrieval, and dense reranking to reduce retrieval failures. Setup is opt-in with built-in privacy gates: run fully on-machine with synthetic prefixes, or opt into Anthropic API or claude CLI tiers for better context generation.

Wiki Ingest automates the process of reading source documents and weaving their content into your wiki. It identifies key entities, claims, and relationships, then creates or updates wiki pages while maintaining bidirectional cross-references and a source index. Each ingest operation is logged for full traceability.

llm-wiki helps you create and maintain a compounding knowledge base as interlinked markdown files, inspired by Andrej Karpathy's pattern. Unlike traditional retrieval systems that rediscover information per query, this skill compiles knowledge once and keeps it current with built-in cross-references and contradiction detection. The agent handles summarization, filing, and maintenance while you curate sources and direct analysis.

This skill guides you through establishing a wiki knowledge base on NanoClaw following Karpathy's LLM Wiki architecture. It walks you through choosing a group, designing your wiki's structure and conventions, creating the necessary directories and schema layer, and configuring how the agent ingests and maintains sources. Each step is idempotent—safe to re-run without clobbering existing work.