Browse / Second brain / LLM wiki / ai-second-brain-vault-template
ai-second-brain-vault-template
hayder-j-ali
Obsidian + Claude Code second-brain template: LLM-maintained wiki, weekly cloud digest delivered as PRs, PR-only main protection on GitHub Free
Open hayder-j-ali/ai-second-brain-vault-template → Checked against the four rules.
claude-codeknowledge-managementllm-wikiobsidiansecond-brain
Works with any agent
ships AGENTS.md, the cross-tool instruction file.
In its own words
A ready-to-clone Obsidian + Claude Code "second brain" for building real expertise, rather than a pile of notes. Claude maintains a compounding wiki for you: you bring sources and judgment, and Claude does the bookkeeping. A weekly cloud agent scouts your field and hands you one worthwhile thing to learn, as a pull request you review.
It's built on the LLM Wiki pattern (a persistent synthesis an LLM maintains on your behalf), plus three things that make it hold up over months:
- A schema Claude follows (
CLAUDE.md): three strict layers (raw sources → wiki → schema), typed pages with frontmatter, an ingest loop where Claude discusses takeaways before writing, a "challenge me" clause, privacy rules for work content, and a self-improvinglearning.md. - A weekly digest that runs in the cloud (
_meta/digest-spec.md): it reads your vault first, tiers new developments by whether they'd change what you believe, keeps a ledger of open decisions, gives exactly one ⭐ action that must pass a "worth your time" test, and pre-drafts an ingest kit. It runs while your laptop is off and arrives as a PR. - PR-only discipline, so nothing lands on
mainwithout your review, even from agents. There are three free layers: a Claude Code hook, a gitpre-pushhook, and a GitHub Action that flags unreviewed changes. That matters because GitHub's branch protection isn't free for private repos.
Quoting 41% of the README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.