The concept library

A concept family is not one skill's output. Three of them build it from different directions, and the tree below is what they have produced so far.

The three directions

concept-family-explorerBreadth

Maps everything around a subject — parent domain, siblings, sub-concepts, adjacent fields, frontier — and finds what is missing from your coverage.

rabbitholeDepth

The narrow inverse: takes one concept and exhausts it — mechanism, edge cases, primary sources, disagreements — until a pass finds nothing new.

llms-concept-abstractorExtraction

Pulls one concept out of a corpus you already hold — the heart from an anatomy textbook, indexing across every database docset — into a concept pack.

They compose: concept-family-explorer finds the node worth the attention, rabbithole exhausts it, and whatever corpus that produces is what llms-concept-abstractor reads back. Breadth saturates when two rounds turn up no new gaps; depth saturates when two passes turn up no new claims. Both stop on evidence rather than on running out of ideas.

The family so far

37 researched concepts under 4 roots, generated 2026-08-30. The full tree is browsable and filterable; the 3D view is the same graph.

• Adult and Membership Ecommerce
  • Adult Ecommerce Operations
  • Adult Product Distribution and Dropship Sourcing
  • Members-Only and Wholesale Pricing
  • Squarespace Commerce and Payments
  • Membership Site Architecture
  • Membership Platforms Build vs Buy
  • High-Risk Payments and Chargeback Management
  • Ecommerce Sales-Tax Nexus Compliance
• Cross-Browser and Mobile Web Programming
  • Cross-Browser Extension and Userscript Portability
  • Browser Engine Differences and Feature Detection
  • Cross-Browser Testing and Debugging
  • Mobile-Web and Browser Security Differences
  • Web-to-Native App Packaging
• Global Context Hub Architecture
  • Concept Tree Tracking
  • Library Registries
  • Execution and Telemetry
  • Code and Prompts
• llms.txt and LLM-readable documentation
  • llms.txt specification v2
  • llms-full.txt page grammars
  • llms.txt generation tooling
  • llms.txt ecosystem evidence
  • llms.txt recreation and family aggregation

Try the extraction step

A bounded, single-pass demo ofllms-concept-abstractor: name a concept, paste up to 4000 characters of corpus, get a mini concept pack. The full skill adds an embedding pass over the whole corpus, several lexicon-expansion rounds, and a verification stage — this shows the keyword-and-classification core.

The public library

No contributed packs yet. Nothing hangs off the public tree that came from outside it, so there is nothing to list — be the first. This space fills from the tree itself: a pack becomes an artifact on the concept it belongs to, and this page reads them back.

How contributing actually works

Contribution is a merge-back proposal against the public tree, not an upload. You take the tree's current sha, build a patch that adds your concept and its artifacts, and submit it toPOST /api/proposals while signed in — a session, not an API key. It goes through a lint gate first: a proposal that fails is still recorded, with its findings attached, so an automatic rejection is reviewable rather than a dead end. A moderator decides the rest.

So the path is: build a pack with the skills above → publish it as your own artifact → propose it onto the tree. Sign in to start; how to read the tree explains what a well-formed node looks like before you propose one.