# second-brain

# second-brain

A personal knowledge base maintained by an LLM following the [LLM Wiki](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f) pattern. Human (bigbes) curates sources and directs exploration; Claude reads, summarizes, cross-references, and keeps the graph consistent.

The vault is designed to be opened directly in Obsidian.

## Layout

```
second-brain/
├── CLAUDE.md           — schema, conventions, workflows (read this first)
├── meta/               — docs about the wiki itself (writing style, tropes to avoid)
├── sources/            — raw immutable source documents
├── inbox/              — Obsidian web clips waiting to be ingested
├── inbox-processed/    — clips after ingest (backup, not deleted)
├── toolbox/            — tool/project pages (type: toolbox)
├── wiki/               — LLM-maintained markdown pages
│   ├── index.md        — content catalog
│   └── log.md          — append-only activity log
└── tools/              — utility scripts (webfetch, etc.)
```

Every file under `wiki/` and `toolbox/` has YAML frontmatter with `type`, `summary`, `tags`, `sources`, `created`, `updated`. Cross-references use `[[wikilinks]]` resolved by Obsidian.

## Workflows

The main entry points live as Claude Code skills under `.claude/skills/`:

- **ingest** — read a new source, create/update pages, cross-reference, log it.
- **query** — look up an answer across the wiki, optionally save as an analysis page.
- **lint** — find contradictions, orphans, stale claims, missing cross-references.

See `CLAUDE.md` for the full schema and `meta/` for style conventions.

## License

Content is personal notes and aggregation. Underlying sources retain their original licenses.