Scraps Ingest implements Karpathy's 'Ingest' primitive to transform fragmented information into atomic, interconnected markdown 'scraps'. It automates the complex labor of knowledge gardening by researching existing wiki state, generating context-aware summaries, and managing link density. The skill identifies relevant topics, handles directory organization, and performs bidirectional cross-linking—automatically updating existing notes to reference new content—ensuring your knowledge base remains a coherent, navigable graph rather than a collection of isolated files.
主な機能
01Context-aware directory organization for title collision management
02Automated content extraction from URLs via WebFetch
03Intelligent bidirectional cross-linking and inbound mention discovery
0445 GitHub stars
05Heuristic-based content density control to protect working memory
06Automated sanity checks for wiki-link and tag consistency
ユースケース
01Summarizing web articles and filing them directly into a personal knowledge base
02Converting LLM-synthesized answers or Q&A results into persistent wiki notes
03Building an interconnected documentation graph for complex software project research