Second Brain OS skill
An Obsidian vault template with 18 skills, 72 commands and 6 subagents for Claude Code: the agent turns sources into linked pages and answers from them
- 1,160 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Writes, renames and merges vault pages, so notes outside git can be lost
- Reads your sources, chats and transcripts, and the scripts work on local vault files
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/second-brain-osDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Writes, renames and merges vault pages, so notes outside git can be lost
- Reads your sources, chats and transcripts, and the scripts work on local vault files
This is third-party code. Review the repository files before installing.
What it does
Second Brain OS is a kit for a «second brain» run by an agent: articles, videos, PDFs and chats become linked wiki pages in plain Markdown files. The repository holds a guide, the `vault-template` starter vault (the `raw`, `wiki`, `projects` and `output` folders, page templates and a `CLAUDE.md`), 18 `second-brain-*` skills, 72 commands in `commands`, six subagents and a few dependency-free Python scripts. The main skill, `second-brain-ingest`, works in steps: it reads the source in full, looks for pages that already exist, writes the source page in `wiki/sources`, extracts concepts and entities, links new pages to old ones in both directions, updates `index.md`, appends to `log.md` and ends with a report of new and updated pages. The author's rule is that a source is not ingested until its page is linked. Contradictions are recorded next to the old version instead of replacing it. `second-brain-query` answers questions from the wiki with links to pages. The other skills cover vault checks (`lint`, `review`, `metrics`, `privacy`), merging and renaming pages, importing chat history and transcripts, preparing reports and publications, and a quiz from your own pages. The scripts compute vault statistics, check links and export the graph.
Who it is for. For people who read and watch a lot and want an agent to keep a personal knowledge base: researchers, engineers, writers and students.
Good fit when
- Bookmarks, screenshots and notes have piled up and you need a way to turn them into a linked knowledge base without filing them by hand
- You want to drop an article, a video transcript or a PDF into `raw/` and get pages with cross-links from the `/ingest` command
- You want to ask questions about your own material and get answers that link to vault pages
- You need scheduled vault checks: broken links, duplicates, unlinked pages, contradictions
- You want a ready `CLAUDE.md` and folder structure so you do not have to invent a vault schema from scratch
Not a fit when
- You need a team wiki: the author says the kit is not for team wikis and does not replace Notion
- You will not add material regularly: the agent will not tidy a vault you never feed
- You want one compact skill instead of a set of 18 skills and 72 commands: karpathy-llm-wiki in the catalog does that
- You only want daily and weekly reviews without source processing: the kit is wider than that job
- Your agent is not Claude Code and does not read the `.claude` folder, where the commands and subagents live, so you would port the skills by hand
Example request
I put an article in raw; ingest it into the wiki: create the source page, extract the concepts and link them to what is already thereLimitations
- MIT licence. The author builds on Andrej Karpathy's 4 April 2026 note on LLM-based wikis and adds the parts the note leaves undefined.
- Installation is manual: clone and copy folders with `cp` into `~/brain/.claude`. On Windows use `python` in place of `python3`: that name often resolves to the Microsoft Store stub and does nothing.
- The agent writes and rewrites your notes, so keep the vault in git. The author advises feeding it ten sources before judging the result: at five pages the graph is not interesting.
- The 72 commands and 18 skills are written for Claude Code. Other agents that read files can use `SKILL.md` and Markdown, but the repository has no ready settings for them.
- The description still has the author's «TODO before publishing» notes: the link to the original article and screenshots are not added yet.
- Part of the `resources` material relies on third-party plugin and repository lists with figures as of September 2026.
How to disable. Delete the `second-brain-*` folders from `~/.claude/skills` and the copied commands and subagents from `.claude/commands` and `.claude/agents` in the vault. Remove the `agents-course` course plugin through Claude Code plugin management. The notes themselves stay plain Markdown files that open in Obsidian without the skills.
FAQ
How do I build a second brain in Obsidian with Claude Code?
Clone the repository, copy `vault-template` to `~/brain` and the skills, commands and subagents into `~/brain/.claude`, then open the folder in Claude Code and as a vault in Obsidian. Put the first source in `raw/` and run `/ingest`. The author advises feeding it ten sources before judging the result.
Which skills come in the kit?
Eighteen, one per workflow: `ingest`, `query`, `lint`, `review`, `merge`, `rename`, `backfill`, `chat-import`, `transcript`, `graph`, `metrics`, `privacy`, `project`, `publish`, `quiz`, `report`, `write` and `changed-my-mind`. The author says `second-brain-ingest` and `second-brain-query` carry the system.
How do skills differ from commands and subagents?
The behaviour lives in the skills, while the 72 commands in `commands` only point at a skill and set its scope, such as `/ingest-pdf` or `/lint`. The six subagents (`curator`, `graph-analyst`, `ingestor`, `linker`, `researcher`, `reviewer`) are separate roles, four of them read-only.
Do I need a vector database to answer questions?
No. The author holds that on a small personal corpus structured pages with links answer better than similarity chunks. The guide has a separate page on the threshold where RAG pays off, and the scripts export the graph for analysis.
What is the agents-course plugin and do I need it?
It is a separate Claude Code plugin with tools from the agents course: a context audit, a goal-test generator, a gate finder, an agent harness audit and an eval-suite bootstrap. A second brain does not need it. Two commands install it: `claude plugin marketplace add undefined-ui/second-brain-os` and `claude plugin install agents-course@second-brain-os`.
What happens when a new source disagrees with an old page?
The skill does not overwrite the page: both positions stay with dates and sources, and the report prints a line about the contradiction found. The `/changed-my-mind` command and the skill of the same name trace how your views changed.
Discussion
No questions or reviews yet.
Related
SkillsFour skills for a personal Obsidian knowledge base: an agent compiles articles and notes into a linked wiki, answers from it and health-checks it
Persistent agent memory in an Obsidian vault: saving decisions, search, research and background vault upkeep
A set of skills and agents for a second brain on markdown and git: notes, a knowledge base and a people CRM with no database or cloud
A skill set that lets an agent run a personal Obsidian wiki: it compiles knowledge from documents and past sessions and answers with citations
An Obsidian vault template with hooks, commands and skills that gives Claude Code, Codex and Gemini CLI persistent memory across sessions
A plugin and skill set that turns an Obsidian vault into a knowledge base with sources, links and grounded answers
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Second Brain OS (this entry) | no | 1,160 | |
| claude-obsidian | no | 15,977 | a plugin and skill set that turn an Obsidian vault into a knowledge base with sources and answers |
| Karpathy LLM Wiki | no | 2,631 | a single skill for a personal wiki: collecting sources, compiling articles with citations and integrity checks |
| Second Brain Starter | no | 851 | a skill that drafts a personal PRD for building a second brain and returns a staged plan |
README in short
The description opens with the problem of bookmarks nobody sorts and a nine-step quickstart. It continues with a flow diagram from `raw` through the agent to `wiki`, a table of repository contents, the guide sections, a seven-module agents course and five handbooks on knowledge graphs, harnesses, loops and evals. It closes with the project's design rules, source links and tables of Obsidian plugins and other implementations of the same idea.
README badge
Original SKILL.md
--- name: second-brain-ingest description: >- Turn raw source material in a second-brain vault into linked wiki pages: 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. --- # Ingest a source 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. ## Core rule Nothing is ingested until it is linked. Every run ends with the new pages connected to existing pages in both directions. ## Workflow 1. **Read the source completely** before writing anything. Partial reads produce pages built from the introduction, which is where sources are least specific. 2. **Check what already exists.** Search the wiki for the main entities and concepts.