# NicholasSpisak/second-brain: an LLM-maintained wiki for Obsidian > This project turns a folder of raw clippings into a structured Obsidian vault using Agent Skills, with ingestion, query and lint commands. It fits people who already read in Obsidian and are willing to run an AI coding agent; the repository does not state a licence. **NicholasSpisak/second-brain** — LLM-maintained personal knowledge base for Obsidian. Based on Andrej Karpathy's LLM Wiki pattern. - Repository: https://github.com/NicholasSpisak/second-brain - Stars: 729 · Forks: 107 - Language: Shell - License: not declared - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/nicholasspisak-second-brain ## The gap second-brain fills between clippings and notes Most note systems fail at the same point. You save twenty articles into a folder, and the folder becomes a landfill. Nothing is summarised, nothing is linked, and searching it means reading it again. second-brain targets exactly that gap: you drop raw material into raw/, an LLM reads it, and the result is a set of structured wiki pages you browse in Obsidian. The README states the division of labour plainly: "The LLM is the librarian. You're the curator." The intended user is narrow. You need Obsidian already installed, an AI coding agent that supports Agent Skills, and Node.js for the npm-based install. If you keep notes in Notion, or if you want a web app with an account and a monthly bill, this repository has nothing for you. It is a local folder, a schema, and four agent skills. ## Raw in, wiki out: the folder schema and what each skill does The architecture is a directory contract rather than a service. The README shows a vault containing raw/ with an assets/ subfolder, a wiki/ tree split into sources/, entities/, concepts/ and synthesis/, plus index.md and log.md, an output/ folder for generated artifacts, and a CLAUDE.md agent config file whose name varies by agent. Four skills operate on that layout. /second-brain runs a guided wizard for a new vault. /second-brain-ingest processes raw sources into wiki pages. /second-brain-query answers questions against the wiki. /second-brain-lint health-checks it. The split matters: ingestion is a write path that produces sources/, entities/, concepts/ and synthesis/ pages, while lint is the repair path for the index and for broken [[wikilinks]]. index.md is described as the master catalog of all pages and log.md as a chronological operation record, which is what lets a later query or lint pass detect drift. The optional tools slot into the read path. qmd is described as a local search engine for markdown files, and the README says the query skill uses it automatically when available. summarize handles links, files and media from the CLI, and agent-browser covers web research. None of the three is required. ## Installing second-brain with npx and running a first ingest The README gives one install command. It requires Node.js and installs the skills into your agent; the README says this covers Claude Code, Codex, Cursor, Gemini CLI and 40+ others. ```bash npx skills add NicholasSpisak/second-brain ``` After that, run the wizard from inside your agent by typing the skill name. The README says it walks you through naming, location, domain and tooling, and that the onboarding script is idempotent, so re-running it will not overwrite existing files. ```bash /second-brain ``` The next step is outside the terminal. Install the Obsidian Web Clipper browser extension and configure it to save into your vault's raw/ folder, then open the vault folder in Obsidian. Clip one article and run the ingest skill. ```bash /second-brain-ingest ``` The README says the LLM will discuss key takeaways with you and build wiki pages. What you should see afterwards is a new page under wiki/sources/ for that source, entries in wiki/index.md, and a line in wiki/log.md. Browse the results by following [[wikilinks]] and opening the graph view. For images, the README's instruction is an Obsidian setting, not a command: set Settings, then Files and links, then Attachment folder path to raw/assets/, and use "Download attachments for current file" after clipping. ## Where second-brain breaks: drift, scale and manual repair The failure modes are documented, and they are all maintenance failures. index.md can fall out of sync with the actual pages. Wikilinks break after a page is renamed. Queries slow down as the wiki grows. The prescribed fix for the first two is the same: run /second-brain-lint, which checks index consistency and offers to fix mismatches, and scans for broken [[wikilinks]] and reports which files need updating. That means correctness depends on you remembering to run lint. The README's own guidance is after every 10 ingests or monthly, whichever comes first, and before major query or synthesis work. The scale answer is to install qmd with npm i -g @tobilu/qmd, after which the query skill uses it automatically. That is a real external dependency for a project whose selling point is a folder of markdown. There is also a recovery boundary. If you modify a file in raw/, the README says the wiki is unaffected because it was built from the original content, and that recovering the original means checking git history if the vault is a git repo, or re-clipping the source. If your vault is not a git repo, that second option is the only one. Nothing in the README describes an undo for a bad ingest beyond deleting the vault and re-running the wizard, which the README lists as the route to a fresh start. It is the wrong tool if you want a knowledge base that stays coherent without periodic manual lint runs, or if you need ingestion to be transactional. ## How second-brain differs from a plain Obsidian vault and from Notion A plain Obsidian vault gives you the same markdown files and the same graph view, and nothing else. You write every page, you place every link, you maintain your own index or you accept that none exists. second-brain adds a fixed schema (sources, entities, concepts, synthesis), a catalog at index.md, an operation record at log.md, and an agent that writes the pages from raw/ input. The trade is that you accept the schema and you accept that an LLM drafts the content. Notion is the other comparison people reach for, and the difference is structural rather than cosmetic. Notion is a hosted workspace with a database model, its own editor, and a subscription. second-brain is a local folder tree that Obsidian opens, and the README lists no price, no account and no server. That also means no sync, no sharing and no mobile story are described in the repository. If collaboration matters more than local files, Notion wins on the axis that second-brain does not address at all. ## Licence, maintenance and the cost of keeping a vault current The repository metadata does not state a licence, and the README does not mention one. If you plan to use this at work, or to build on the skills, that is the first thing to resolve: check whether a LICENSE file exists in the repository before you depend on it. Nothing here is legal advice, but an unstated licence is a real adoption blocker for a company. On maintenance, the last push to the default branch was on 2026-04-07. The repository is not archived. There are no releases listed, so there is no versioned upgrade path to follow; you track the main branch. The upgrade cost is therefore the cost of re-reading the skills after a change, plus whatever the wizard regenerates in your agent config files. The README notes that the wizard generates config files for each agent you select and that they all follow the same wiki schema, which is what allows multiple agents to work on the same vault. The recurring cost is your own time. Lint after every 10 ingests or monthly, plus before major query work, is a schedule you have to keep. qmd is an extra global npm package once the wiki grows. Neither cost appears in a billing page, which is precisely why it is easy to underestimate. ## Conclusion Adopt it if you already read and clip in Obsidian and run Claude Code, Codex, Cursor, Gemini CLI or another Agent Skills host, because the four skills and the raw/wiki/output layout are the whole product. Do not adopt it if you want a hosted service, a fixed monthly price, or a wiki that stays correct without you running /second-brain-lint. Before committing, check whether a LICENSE file exists in the repository, confirm your agent appears in the wizard's list, and run one ingest on a single clipped article to see whether the generated pages match how you file notes. ## FAQ ### How do I install second-brain? Run npx skills add NicholasSpisak/second-brain with Node.js installed. The README says this installs four skills into your AI agent, covering Claude Code, Codex, Cursor, Gemini CLI and 40+ others. ### How do I set up second-brain with Claude? Install the skills, then type /second-brain in your agent. The README says the wizard walks you through naming, location, domain and tooling, and that it is idempotent and will not overwrite existing files. ### How do I use second-brain with Obsidian? Launch Obsidian, choose "Open folder as vault" and select your vault folder. Clip sources into raw/ with the Obsidian Web Clipper, run /second-brain-ingest, then browse wiki/index.md and follow the [[wikilinks]]. ### How do I use an AI as a second brain with this project? The agent does the writing and you do the curating. The README describes the LLM as the librarian: it reads raw/ material, writes wiki pages, creates cross-references and maintains the index, while you decide what goes in. ### What does second-brain cost per month? The README lists no price, account or subscription. The only costs it names are the optional tools, installed as npm packages such as npm i -g @tobilu/qmd, plus whatever your AI agent charges. ## Sources - [Issues](https://github.com/NicholasSpisak/second-brain/issues) - [NicholasSpisak/second-brain on GitHub](https://github.com/NicholasSpisak/second-brain) - [README](https://github.com/NicholasSpisak/second-brain/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/nicholasspisak-second-brain