Everything an agent needs is inside one repository, and each file points to the next.
Five reads, then a screen.
Map, rules, catalogue, one component, compose. Nothing is shown before it is needed; nothing needed is left unsaid.
What is actually there.
Both are inside the line.
5 built, 8 pending.
A skill routes a recurring job to the right files. Six of the ten indexed systems ship them; the top-scoring one ships a skill that sends the agent to a page template instead of inventing chrome. Every skill here is checked against the files, tokens, exports and scripts it names, so one that goes stale fails the build.
| Skill | Does | State |
|---|---|---|
| pick-a-page-template | What page type is this, then the shape, then the not-built list. | built |
| use-a-token | The token for the job and the utility that carries it, never a raw value. | built |
| add-a-component | The five files, the catalogue row, the barrel export, the gates. | built |
| check-adherence | Run the gate, read the triage, resolve a gap honestly. | built |
| compose-a-screen | Page, header, sections, stacks, grids, then the six questions. | built |
| review-a-screen | answer the six composition questions against a render | pending |
| add-a-token | propose a token, get the ruling, wire the utility | pending |
| write-a-component-doc | the COMPONENT.md shape, for and not-for first | pending |
| choose-a-status-role | success, warning, danger, info, and what each means | pending |
| handle-a-gap | render Missing, declare it, let triage rank it | pending |
| run-the-gate | undrift gate before a commit, read the fix it names | pending |
| theme-a-screen | light and dark from one source, never a dark: utility | pending |
| build-a-list-page | the list template when its kit exists | pending |
The index measures ten public systems against this shape. The thesis says why the shape is the one that holds.