Starter Cairns Packs.
Three pre-built Cairns knowledge bases that ship with the workshop. Research papers (52 AI papers with citation graph), tool docs (10 services including Claude Code, Slack, Supabase, Vercel), and YouTube (a curated transcript corpus). Borrow any of them as-is to see how Cairns work in practice, or use them as starter material when you build your own.
- Three knowledge bases ready to query the moment you finish onboarding. No corpus-building required.
- A worked example of what a finished Cairns vault looks like at scale. Useful before you build your own.
- A query practice ground for the librarian, ingest, lint, and dream skills. Try the tools on real data instead of waiting until your own vault is full.
How this fits in the OS.
You have the workshop repo cloned and the Cairns skills installed (the /personalize flow installs them, or you can install them manually).
Whichever pack you borrow. Each one is self-contained: a vault folder, manifests, and example pipeline scripts.
A working knowledge base on day one. You can query the curated material with the same skills (/cairns-query, /cairns-lint, /cairns-dream) you will eventually use on your own vault.
What ships in the box.
The workshop repo includes three pre-built Cairns knowledge bases under cairns/. Each one has the same three-layer structure: L1 waypoints, Chain-of-Density cards in the L2 Chain of Density Card Catalog Layer (Card Catalog Layer for short), L3 raw evidence. The difference is the source material.
An AI research paper corpus with the citation graph as the differentiator. Each paper is an L3 raw document. Each gets a Chain-of-Density card in the L2 Chain of Density Card Catalog Layer. Each contributes to one of several L1 waypoints (alignment, retrieval, agentic patterns, training methods, the rest). The graph holds :Paper nodes and :CITES relationships, so you can ask citation-lineage questions and get answers.
When to reach for it. When a question depends on which paper introduced an idea, what later work built on it, or what the dependency chain of a method looks like. The graph beats vector search on questions like "what method does X depend on?"
Lives at cairns/research-papers/. The README in that folder has the rebuild instructions if you want to refresh the corpus.
A Cairns over the workshop's tool stack documentation. Ten services covered: Claude Code, Anthropic SDKs, Cursor CLI, n8n, the Slack API (the big one, 311 source cards), Supabase, Vercel, Bland CLI, the Telegram Bot API, and the 1Password CLI. Each tool gets its own L1 Cairn; each section of the docs becomes an L2 Chain of Density Card Catalog Layer card; the original documentation is preserved as L3.
When to reach for it. When you are building anything that touches one of these ten services and want fast, accurate, narrow answers about endpoints, parameters, or flows. It is dramatically better than copy-pasting from the live docs into Claude.
Lives at cairns/tool-docs/.
A Cairns over a curated YouTube transcript corpus. The same three-layer structure with transcripts as L3, Chain-of-Density cards in the L2 Chain of Density Card Catalog Layer, and topic waypoints as L1. Plus a graph export with entity links across videos so you can ask cross-source questions ("what does the team say about X?").
When to reach for it. When you want a worked example of the YouTube ingest pipeline before building your own, or when you want to query a body of video content for a specific concept without watching every video.
Lives at cairns/youtube/.
Borrow them, or use them as scaffolding.
Two ways to use a starter pack:
- Borrow as-is. Query the pack with the standard Cairns skills.
/cairns-queryworks on it./cairns-lintaudits it./cairns-dreamproposes new L1 waypoints. You learn the rhythm of working with Cairns before you ever build your own. - Use as scaffolding. Each pack is a worked example of the same architecture you will use for your own vault. Open
cairns/research-papers/vault/cairns/L1/_index.mdto see what an L1 index looks like at scale. Opencairns/tool-docs/vault/card-catalog/L2/to see what mature Chain-of-Density cards look like. Steal the patterns; replace the content.
The agent does this for you. Each pack ships with a pipeline/ folder containing the scripts that built it. If you want to rebuild any pack from source (refresh the papers, re-scrape the docs, re-ingest the videos), the README in each pack's folder tells you which commands to run.
How to query a pack.
The Cairns skills work the same way on every pack. The vault path is the only thing that changes:
/cairns-query --vault cairns/tool-docs --question "what are the rate limits on the Slack API?"
Claude reads L1 first (waypoint), drills to L2 Chain of Density Card Catalog Layer (the relevant card), and only opens L3 (raw source documentation) when it needs exact detail. The same query against the research-papers pack would walk the citation graph if the question is about paper lineage.
If a pack is missing something you wish it had, you can extend it. The starter pack is your base; your own ingestions layer on top.
When a starter pack is the wrong move.
These packs are general-purpose. They are not personalized to your business, your clients, or your specific work. If you need a Cairns over your meeting transcripts, your client decisions, your own writing, or your industry-specific research, the starter packs are not the answer. They are a teaching ground.
The natural progression: borrow a starter pack for a week, get fluent with the query patterns, then build your own vault on top of your real material. The Cairns Three Tiers and Cairns Ingest Pipeline cards are the next steps.
Where to go next.
Three natural next moves depending on what you want to do: