second-brain-ingest

A knowledge-management skill that processes raw source documents into structured, interlinked wiki pages for a personal second brain. It activates when a user adds files to a raw/ directory and wants them ingested, or says things like process this source, ingest this article, or I added something to raw/. It first identifies which files need processing, either using files the user names or detecting unprocessed files by listing raw/ (excluding raw/assets/) and comparing against ingest entries already recorded in wiki/log.md.

For each source it follows a defined workflow: read the source completely (noting and separately reading important images), then share the three to five most important takeaways and wait for user confirmation before writing. It creates a factual source summary page under wiki/sources/ with YAML frontmatter and sections for summary, key claims, entities mentioned, and concepts covered. For every entity such as people, organizations, products, or tools, and every concept such as ideas, frameworks, theories, or patterns, it either updates an existing page in wiki/entities/ or wiki/concepts/ (adding the source to frontmatter, updating the date, and noting contradictions with both sources cited) or creates a new focused page. It then wires related pages together with wikilink references, updates wiki/index.md with one-line summaries, appends an ingest entry to wiki/log.md, and reports what was created, updated, and found.

Its conventions keep the knowledge base coherent: source summary pages stay factual, with interpretation reserved for concept and synthesis pages; a single source typically touches ten to fifteen wiki pages; existing pages are preferred over new ones; and internal links always use wikilinks rather than raw file paths. The skill is allowed the Bash, Read, Write, Edit, Glob, and Grep tools. After ingesting, the user can query the knowledge base with a companion query command, ingest more sources, or run a lint health-check after every ten ingests.