>- read the source, split it into concepts and entities, write or update pages, connect them to existing pages, and record the run in the log. Use this skill whenever the user drops a file into raw/, pastes an article, transcript or PDF and asks to add it to the vault, says "ingest this", "add this to my second brain", "process raw", or asks to catch up on unprocessed sources, even if they do not name the ingest command. Do NOT use for answering questions from an existing vault, for linting or repairing pages, or for editing notes the user wrote by hand.
npx skills add https://github.com/undefined-ui/second-brain-os --skill second-brain-ingest
An ingest that only writes a summary page produces a vault that grows without
getting smarter. The whole value is in the second half of the job: connecting
the new material to what is already there. A page that lands unlinked is
invisible within a week.
Nothing is ingested until it is linked. Every run ends with the new pages
connected to existing pages in both directions.
produce pages built from the introduction, which is where sources are least
specific.
concepts. Ingesting into an empty vault and ingesting into a vault of 300
pages are different jobs: in the second, most of your work is updating
pages, not creating them.
wiki/sources/. Record claims as claims, withthe source attached.
exists, add what this source contributes and update updated:. If it
contradicts what is there, record both positions rather than replacing.
entity it touches; each of those links back. Add any missing target pages,
or list them under Gaps in index.md.
index.md and append to log.md in the same run, because anindex that lags is how a vault starts drifting.
End every run with this exact shape:
Ingested: <source name>
New pages: <n> (list)
Updated pages: <n> (list)
Links added: <n>
Contradictions found: <none | description>
Gaps created: <list of linked pages that do not exist yet>
Over-extraction is the common failure: ten thin concept pages from one article,
each a restated paragraph. A source usually yields one to three concepts worth
their own page. If a candidate concept cannot be explained without referring
back to this one source, it belongs inside the source page instead.
Under-extraction shows up as source pages that contain three separate ideas and
link to nothing. If a page needs section headings for unrelated topics, split
it.
Input: the user drops raw/karpathy-llm-wiki.md, a clipped gist, into a vault
that already has pages for [[Obsidian]] and [[Claude Code]].
Output: one source page; a new concept page [[LLM wiki]] explaining the
pattern in plain language; an entity page [[Andrej Karpathy]]; updates to
[[Obsidian]] and [[Claude Code]] noting their role in the pattern; a link
from the existing [[Personal knowledge management]] concept to the new one;
index and log updated; report showing 3 new, 3 updated, 11 links.
Take undefined-ui/second-brain-ingest from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.