The wskill projects into two views. The book you're reading is one; the other is a skill — the :ai_skill folder wcl wdoc skill builds, a SKILL.md plus a references/ directory an agent loads. SKILL.md inlines the curated :ai/:both indexes, linking each pinned note's page directly. This page maps that skill, then shows each page's actual generated Markdown.

Each circle is a skill page — SKILL.md purple, the per-note pages green. Each line is a link the skill actually emits. Scroll to zoom, drag to pan.

Force-directed graph of the skill's pages and the links between them.SKILL.mdSeparationof Data andPresentationSelf-ContainedContentAtomicNoteDecomposinginformationStructureddataAssets(images &datafiles)ConceptEntityFactProcessIndexWhat isit?The viewfamilyAudiencecontrolVisibilityscoping(@only /@except)Anatomyof the AIskillThepresentationviewThetrainingviewCustomprojections(schemaextensionmodules)Components:one lookfor everyunitCollectionsandregistriesClaudeCodeWhich unitkind? — thedecisionguidePlain skillor wskill? —the routingruleThewskillfolderlayoutThe wclcommandsan authorusesThestandardjustrecipesCreatinga newwskillAddingcontentto awskillCapturingresearchinto awskillBuildingthewskillindexSettingup AIskillgenerationBuildingandinstallingthe AIskillAttachinga wskillto aregistryCreatinga schemaextensionCreating thepresentationviewCreatingthetrainingviewUpdating awskillwhen itssourcechangesUpgradinga wskillto a newbaseschemaReviewinga wskill(human ⇄agentloop)Editing awskill inthebrowserCanonicalClaude-skillformat(meta-skill)

§ 1Pages

Every file the skill emits has its own preview page — open one to read its real generated Markdown (the same output wcl wdoc skill writes, internal links and all) and comment on any line (wcl wdoc serve --comment) to flag a fix. Clicking a circle in the graph above opens the same page.

§ 1.1Entry page

SKILL.md — the entry point; it inlines the curated indexes and links each note's page directly.

§ 1.2Process pages

NoteBook page
Creating a new wskillopen
Adding content to a wskillopen
Capturing research into a wskillopen
Building the wskill indexopen
Setting up AI skill generationopen
Building and installing the AI skillopen
Attaching a wskill to a registryopen
Creating a schema extensionopen
Creating the presentation viewopen
Creating the training viewopen
Updating a wskill when its source changesopen
Upgrading a wskill to a new base schemaopen
Reviewing a wskill (human ⇄ agent loop)open
Editing a wskill in the browseropen

§ 1.3Concept pages

NoteBook page
Separation of Data and Presentationopen
Self-Contained Contentopen
Atomic Noteopen
Decomposing informationopen
Structured dataopen
Assets (images & data files)open
Conceptopen
Entityopen
Factopen
Processopen
Indexopen
What is it?open
The view familyopen
Audience controlopen
Visibility scoping (@only / @except)open
Anatomy of the AI skillopen
The presentation viewopen
The training viewopen
Custom projections (schema extension modules)open
Components: one look for every unitopen
Collections and registriesopen

§ 1.4Entity pages

NoteBook page
Claude Codeopen

§ 1.5Fact pages

NoteBook page
Which unit kind? — the decision guideopen
Plain skill or wskill? — the routing ruleopen
The wskill folder layoutopen
The wcl commands an author usesopen
The standard just recipesopen

§ 1.6Research pages

Captured findings — shipped at references/research_<id>.md (plus the index_research menu) so external consumers can discover them mechanically.

FindingChecked
Canonical Claude-skill format (meta-skill)2026-07-05