Five folders. Flat.
Every note is one search away — for you and your agent. No nested hierarchies to maintain. Links and metadata do the organizing.
Obsidian vault template + agent skill
An Obsidian vault your agent can actually read.
Your notes become searchable context for your AI agent - not dead files in a folder. Every conversation picks up where the last one left off. One command from empty folder to working system.
400+ notes → 0 seen by your agent
Weekly captures → Never processed
Every conversation → Starts from scratch
Agent reads. Human writes.
The vault is your externalized thinking. The agent is a reader, navigator, and synthesizer — never a ghostwriter.
Agent reads
- Searches across your vault
- Surfaces forgotten connections
- Finds contradictions across time
- Suggests links and tags
You write
- One atomic idea per note
- Titled as complete, declarative phrases
- Dense
[[wikilinks]], stubs encouraged - Your thinking, not summaries
Everything you need.
Nothing you don't.
Typed frontmatter
note, doc, journal, article, book, youtube, tweet, and more. The agent knows what each note is and can filter, group, and reason about your vault by type.
Clipper → Inbox → Notes
Clip any page. The template auto-classifies by URL and queues it. Process when you're ready — keep, convert to a note, or discard. Nothing rots.
Bases, not Dataview
First-party YAML views the agent can create and edit. No fragile query language. Ships with an inbox triage view; add reading queues and watch lists as you go.
QMD hybrid search
Ask a question, get your own notes back — ranked by relevance. Keyword, semantic, and hybrid search across collections. Ripgrep fallback when QMD isn't installed.
Skill-based context
The agent learns your vault conventions from a bundled skill — not a file you maintain. OpenClaw memory paths are wired at init so context carries across sessions.
What it looks like inside.
Zettelclaw — Obsidian
One command. Done.
Want to verify your setup? npx zettelclaw verify