<x-form> renders data-md-form and data-md-form-actions, and its column
and actions row move to components/form.css in material.components, on
the spacing tokens (plan step 36). The actions slot's own attributes
still land on the row.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
A rewritten component's padding, margin and border sit in
material.components, which the Workbench declared below all of
Tailwind's layers, so preflight's universal reset beat them. The entry's
layer statement now puts material between base and components (plan
step 36): the material reset already carries preflight, and every
utility still outranks the package.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
tailwind.css and components.css now hold one block per audit group, and the new
layout.css gathers the layout components, so the parallel rewrites of plan steps
35 and 36 each edit their own lines. .browser-lock is the lock that keeps those
streams to one browser at a time.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
<x-icon size> writes --md-icon-size (24px by default) and takes the optical-size-20
cut for 20px or less, as M3 asks; mirror-rtl flips a directional symbol. <x-shape
size> writes --md-shape-size. Both draw from plain-CSS stylesheets in
material.components, gathered in components.css until all.css replaces it; a
caller's size class still wins while the other components are rewritten.
Plan steps 35 and 36, the prop every component rewrite depends on.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan steps 33 and 34. The Tokens section gains a paragraph naming
foundation.css as the one required import, its layer order, and the
md-* text classes as 2.0.0's vocabulary; the full rewrite of the docs
is step 47.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan steps 33, 34 and 42. StylesheetsTest follows foundation.css through
its imports and checks each file: a header, the exact layer statement,
plain imports only before its first block, every rule in its own
material layer but the two hiding rules, no Tailwind directive, media
queries only at 600/840/1200/1600px. It also pins the spacing tokens to
the reference table, the md-type classes to the type utilities'
declarations, the whole text class set, the interaction classes against
their utilities, one [x-cloak], material.css never reaching the reset,
and the Workbench importing the foundation before Tailwind.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 33. The Workbench imports foundation.css ahead of
@import 'tailwindcss', so the material layers are declared first and
sit below Tailwind's, then tailwind.css: the theme, the token utilities
and the component stylesheets still written for Tailwind, with no
tokens, so nothing is declared twice. material.css keeps working for
1.x-style imports after Tailwind by bringing the tokens, the hiding
rules and tailwind.css, and never the reset, which would outrank
Tailwind's utilities from there.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 34. text.css (material.text, imported by the foundation) is
the fixed set for text on plain elements: md-type-* and
md-type-emphasized-* with the type utilities' declarations, the role
inks md-ink to md-ink-inverse, md-text-start|center|end, md-truncate,
md-line-clamp-2|3, md-nowrap, md-tabular and md-visually-hidden. Its
header says when to use each, from the material-3 guideline and
docs/reference/m3/styles.md § Typography.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 33. foundation.css is the one required import: the full
material layer statement, then plain imports of its parts, each opening
with the same statement.
- foundation/reset.css (material.reset): Tailwind 4's preflight rule for
rule, with M3's on-surface-variant placeholder and no typeface.
- foundation/hidden.css: [hidden] and [x-cloak], outside every layer.
- foundation/tokens.css (material.tokens): the token files.
- foundation/base.css (material.base): the page's surface, ink and
brand typeface, and the font face.
- foundation/interaction.css (material.base): md-state-layer,
md-focus-ring, md-touch-target and md-link, the plain-CSS forms of the
utilities; md-state-layer also takes data-md-dragged, and
md-touch-target is 48px (space600) where the utility reads 3rem.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 33. tokens/spacing.css declares --md-sys-measurement-space25
to space900 (2 to 72px) in material.tokens, from
docs/reference/m3/styles-supplement.md § Spacing, in px because a dp
is a CSS pixel and spacing does not scale with text. The design skill
names the new token family, as its drift test requires.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 33. Each token file now opens with the material layer
statement and wraps its declarations in material.tokens (the font face
in material.base), so an application's unlayered scheme outranks the
package default whatever the selectors, and the foundation can import
the files with no layer() on the import.
The type-* and state utilities move to tokens/utilities.css, still
imported by material.css, so no file the foundation imports carries a
Tailwind directive. [x-cloak] moves beside the [hidden] rule in
foundation/hidden.css, outside every layer; material.css imports it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. The chip set's scroll buttons (plan step 24) moved
`x-data="materialChipSet"` from the row to a wrapper around the row and
its buttons, so `[data-chip-set][x-data]` found the first wrapping set on
the page, which never overflows.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. inputs.md IN-01 moved `aria-expanded` from the raw input to
the `role="combobox"` wrapper around it, and IN-09 put a scrim under the
docked view, so a press outside lands on the scrim and Playwright could
never press the text beneath it. The tests read the wrapper and press
the scrim.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. The Phase C rewrite checked the stored collapse at 1600px
with a new visit(), but every visit opens a fresh browser context with
empty storage, where the rail rightly starts expanded. The test now
resizes the page that holds the choice.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. The full-screen range picker (plan step 24) added a second
`[data-datepicker-headline]` in its app bar, hidden unless the picker is
full screen, so Playwright's strict locator found two headlines in every
modal picker. The tests read the modal header's.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. Phase E put a standard side sheet (containment.md, step 23)
above the modal one on the showcase page, so the first `[data-sheet]` scrim
the test pressed belonged to it; the test now presses the modal sheet's
own scrim. The preset-heights bottom sheet's button made "Bottom sheet"
match two buttons under Playwright's strict locators; it is matched by its
exact label. The actionable card test chained an assertion on script(),
which returns the script's value rather than the page.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. Since actions.md ACT-02 the host is the permanent polite live
region, in the page before any message and after the last, so the test
that waited for it to disappear once Undo ran could never pass. It now
waits for the toast to go and checks the host stays.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32, found by the Browser suite. Content padding (16dp by default
since containment.md C-12) shrinks every keyline near the ends, so large
items resting at the start or the end carry a small inset. Bring-into-view
took any inset over half a pixel for "not fully open", so focusing or
pressing a large item at rest scrolled the row to it. The focal keyline's
own inset at the current scroll is the threshold now. The press test also
checks that focusing and pressing an open item moves nothing.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32, found by the Browser suite. Since containment.md C-18 put
the tab stop on each item, the item itself matches the selector for
interactive controls, so every press on an item read as a press on a
control inside it and never scrolled a partly hidden item into focus. A
mouse press still worked through the focus it gives the item; an activation
that clicks without focusing, as a test or assistive technology does, did
nothing. Only a control other than the item counts now.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 32. Since containment.md C-12 the carousel defaults to M3's 16dp
of content padding, and Compose's createShiftedKeylineListForContentPadding
shrinks every keyline near the ends by a share of it, so a large item at
rest at the start or the end is masked by 1.33px a side in the showcase.
The tests asked for a zero inset there and failed; `open(i)` asks whether
an item is at the largest size any item has at that position. The helper's
list of showcase carousels gains the multi-aspect one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 22, item 8 (actions.md § Missing, "Adaptive menu → bottom sheet
at compact"): M3 says "at compact breakpoints, consider swapping a menu for
a bottom sheet", and nothing connected <x-menu> to <x-bottom-sheet>.
<x-menu sheet-at-compact> draws its slot twice, in the popover and in a
modal <x-bottom-sheet> teleported to <body>, and below `medium`
(upTo('medium')) the trigger opens the sheet. The trigger says
aria-haspopup="dialog" there and "menu" from medium, aria-expanded in
both. Items keep their roles and the APG keyboard in the sheet; choosing
one, Escape or Tab close it and return focus to the trigger; a submenu
opens in place under its item; `filter` works in both; a resize across
600px closes whichever is open. The sheet has a fixed id that menu.js
makes unique and keeps as its wire:key, and the lists are keyed, so a
Livewire render patches an open sheet instead of swapping it.
bottom-sheet.blade.php is used as is. The filter now finds its field and
empty row per list instead of through x-refs.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Text fields bounded from medium, dialog scroll dividers and the new meaning of
separator, the two-line snackbar, bottom sheet stops, chip set scroll buttons,
and a list of the configurations 2.0.0 adds.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Chrome makes a scroll container with nothing focusable inside a keyboard stop, and
showModal() gives it the dialog's first focus, which showed the browser's own ring.
The body now takes M3's 3px secondary indicator, inset so the dialog's rounded,
overflow-hidden box does not clip it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
The condensed reference said a docked toolbar is never rounded and left 'it'
ambiguous for the large-screen form; the guidelines page says straight corners,
except that on web and large screens the docked toolbar can be rounded.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 23 (containment). The containment showcase gets a dialog with a
long body, whose rules follow its scroll, and one with `separator`, whose
rules always show. The <x-modal> entry in SKILL.md says what `separator`
means now and that every dialog divides a scrolling body by itself.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 23 (containment), containment.md § Missing: "no divider pinned
between a scrolling body and the header/actions". Every dialog now draws a
1px outline-variant rule under the header once the body is scrolled away
from its top, and over the actions while more of it is below; a body that
fits shows neither. The rules are pseudo-elements in the rows' own padding
(dialog.css), so showing one moves nothing, and x-dialog-dividers
(dialog.js) marks the wire:ignore.self <dialog>, which a morph leaves
alone, watching scroll and a ResizeObserver on the body and a wrapper
around the slot. The M3 gaps are split around the rules (8/8 under the
header, 8/16 over the actions). A full-screen dialog's rule sits under its
phone bar, and its action bar's always-on border follows the scroll too.
`separator` now means "draw both rules always" instead of rendering two
<x-divider> elements. The fade uses the effects-fast token, zero under
reduced motion.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 25, item 8 (navigation Missing: "Toolbar, large-screen
treatments"). M3's toolbar guidelines (Adaptive design, Resizing) say it
of the docked toolbar: "On web and large screens, the docked toolbar can
be rounded. Dividers can be used to organize large amounts of items",
and it can be placed in different parts of the page. `rounded` does that
from expanded (840px): fully rounded, spanning its container, lifted
16px off the window's edges at place="bottom"; below expanded it stays
the square full-width bar. An <x-divider vertical /> between groups of
controls stands as tall as the icon buttons inside any toolbar. The
floating toolbar is already fully rounded and gets no new form.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 25, item 7 (navigation Missing: "App bar, trailing-action
overflow at small widths"). `<x-app-bar :actions="[...]">` takes the
trailing actions as a list whose entries take <x-menu-item>'s props plus
any action attribute. M3 allows up to two trailing icon buttons, and up
to four on larger screens, and lets trailing actions collapse into an
overflow menu at smaller breakpoints: below medium the bar keeps at most
two icon buttons, from medium four, the "More options" button counted
among them, the rest in its menu. Both forms render and media queries
pick one, so there is no script and no flash; each width has its own
menu so the arrow keys never reach a hidden row. The actions slot still
takes markup and never overflows.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 25 (navigation Missing): M3's toolbar configuration table lists
"With FAB" for both toolbars, but the `fab` slot was gated to the floating one.
A docked toolbar now takes it at its end, inside the toolbar where the arrow
keys reach it; its controls gather at the start, and the FAB rests flat on the
bar — elevation 0 for a FAB nested in another component, as Compose's
bottomAppBarFabElevation is.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 22, actions.md § Missing (FAB to extended FAB scroll collapse):
M3 has an extended FAB collapse to a FAB on scroll-down and re-extend on
scroll-up, and `<x-button fab>` only swapped on window width.
`<x-fab collapse-on-scroll>` (extended, with an icon) takes a
`materialFab` flag from the new resources/js/fab.js, which reads the
window's scroll once a frame, ignores moves under 8px and extends again
near the top. The morph is CSS in actions.css: the label sits in a grid
track that closes to zero, the gap and minimum width follow on the
default spatial spring, and the label fades on the effects one; reduced
motion swaps outright through the motion tokens. The label is clipped, not
removed, so the collapsed FAB keeps its accessible name.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 23, containment.md § Missing ("Divider: the divider-with-text /
subheader configuration"). `<x-divider text="…">` draws the label at the
start and the rule running on from it, with the divider specs table's
geometry: 4dp between the words and the rule, 8dp right margin, 8dp
bottom margin. The label is M3's subhead — title-small in
on-surface-variant, the only subhead type M3 publishes (the rich
tooltip's specs), since the divider page names none. The words stay text
naming the group; only the rule is the separator.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 25 (navigation Missing): M3's "Expanded behavior | Hide when
collapsed" was only implicit below `medium` in the adaptive rail.
`<x-navigation-rail hide-when-collapsed>` (collapsible and adaptive rails) takes
the rail out of the layout when it is collapsed; `$store.rail.show()` from an
app-bar menu button brings it back expanded over a scrim, sliding in, and its
own menu button docks it again. It does not reach the bands where the window,
not the visitor, collapses a rail (below `medium` for collapsible, `medium` for
adaptive), since M3's collapsed rail may never hide. Every `rail-collapsed`
branch now stops while the rail is open. `<x-app-shell hide-rail-when-collapsed>`
exposes it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 23, containment.md § Missing ("Carousel: the uncontained
multi-aspect-ratio layout"). `<x-carousel layout="multi-aspect">` is the
layout M3 added in November 2025: each `<x-carousel-item aspect="…">`
keeps its own ratio at the row's fixed height, held inside M3's 9:16 to
16:9 range (a square by default), with 16dp leading padding, 8dp gaps,
the extra-large corner and uncontained (default) scrolling.
The keyline machinery does not fit it: an Arrangement counts items of
one size each, and every snap position and mask follows from that size.
So the layout is a plain flex row, unmasked, and carousel.js only
measures each item's resting position off the DOM, which keeps the
previous/next buttons, the arrow keys, Home/End and bring-into-view
working. The header and SKILL.md say so.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 22, actions.md § Missing (two-line snackbar height, snackbar
keyboard shortcut): the container grew organically from `min-h-12` rather
than sitting at SnackbarTokens.TwoLinesContainerHeight, and there was no
way for a keyboard to reach an actioned snackbar at all.
A `description` is M3's second line, so the snackbar is pinned to 68px
whenever one is there, and below `medium` a two-line snackbar with an
action wraps the action under the text — M3's "two lines with longer
action" configuration. Alt+G, the shortcut M3 suggests for the web, moves
the focus to a snackbar that carries an action from wherever the page had
it; `event.code`, because Alt rewrites `event.key` on some layouts.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
M3 § Chips, Accessibility asks that a row which overflows horizontally carry a
visible affordance; `<x-chip-set scroll>` faded its edges and left it there
(plan step 24, audit docs/audits/m3-alignment/inputs.md § Missing). Where the
pointer is fine, and so there is no swipe to reach for, a button now sits over
each fading edge and scrolls the row by most of its width. They are pointer
affordances only — no tab stops, since the arrow keys already walk every chip —
and the row's scroll padding grows on a fine pointer so a chip the keyboard
reaches clears the buttons as well as the fade. The scroller became a ref,
because the buttons stand outside it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 22, actions.md § Missing (Selection-required / multi-select
semantics): M3 lists single-select, multi-select and selection-required as
button-group configurations, and `<x-button-group connected>` left
selection entirely to the caller's own `aria-pressed`.
`selection="single|multi"`, with `required`, takes `aria-pressed` over:
pressing a button writes its `value` to `wire:model` or `x-model`,
deselects the others in `single`, and refuses the press that would leave
nothing selected. With no model it reads the buttons' own `aria-pressed`
once and goes on from there.
`<x-group>` and this do not absorb one another, and the header and
SKILL.md say why: `<x-group>` is for a choice whose options are data — it
renders real inputs, posts in a plain form and paints its own segments —
and stays the first thing to reach for; a selection group governs buttons
the caller writes, manages state and shape, and leaves each button to draw
its own colours from its own `:selected`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 23, containment.md § Missing ("Bottom sheets: preset heights
and the handle's cycle-through-heights behaviour"). `heights` (or `snap`,
which is 25/50/90dvh) lists the stops a sheet moves between; it then
takes its stop's height and opens at the stop that equals `height`, or at
the first. Fewer than two stops is no stops, since M3 only requires the
non-drag alternative "if multiple preset heights exist".
That alternative is the drag handle, which M3 names a button: activating
it moves to the next stop and announces it in a live region, and from the
last stop it closes the sheet — M3's "selecting the drag handle toggles
preset heights or closes the sheet", and the same thing a handle with no
stops has always done. A drag now runs the sheet's height with the
pointer and settles on the nearest stop, or closes below the smallest one
or on a downward flick.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
M3 § Search names three entry points and only the bar existed; suggestions
before the first keystroke were missing too (plan step 24, audit
docs/audits/m3-alignment/inputs.md § Missing). `trigger="icon"` is the search
icon button — search as a secondary action, one 48px button that expands into
the full-screen view at any width, since an icon button has nothing to dock
under, and takes its focus back on close. The `suggestions` slot stands where
the results do until something is typed; the live region counts whichever of
the two is on screen and names suggestions as suggestions.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 25 (navigation Missing): M3's "alignment: top or center; on tablets
prefer center alignment for reach" had no implementation — the rail was
top-aligned only. `<x-navigation-rail align="center">` centres the destinations
between the header and the footer, which stay at their ends because M3 keeps the
menu icon and the FAB top-aligned. `safe center` gives the top back the moment
there are more destinations than fit, where centring would push the first one
out of reach above the scroller.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 22, actions.md § Missing (Square-by-default button groups): M3
lists "Default shape | Round, square" as a button-group configuration and
neither `<x-button-group>` nor `<x-group>` had a `shape` prop, so a group
of square buttons still drew round outer corners.
`shape="square"` passes the square corner scale to every child through
`data-shape` and two new variables in groups.css: `--group-outer` is what
the two ends of a group round to (half the height when round, the corner
its inner edges take when square — M3's square table, 4/8/8/16/20dp), and
`--group-square`/`--group-square-pressed` give a standard group's buttons
the scale `<x-button shape="square">` draws. The selected rules keep
`--group-full`, so a selected button in a square group rounds: M3 has a
toggle inside a group swap shape on selection either way.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 25 (navigation Missing): NavigationRailCollapsedTokens
.NarrowContainerWidth, 80dp, had no way in — only the default 96.
`<x-navigation-rail width="narrow">` takes it; the collapsed width is now a
variable, so the four rules that each have their own reason to name it follow
without a specificity fight. No label fits under a 56px indicator at 80px, so a
narrow rail's items are their icons alone, with the label taken out of the
drawing and left in the accessibility tree, where it names the destination.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Plan step 23, containment.md § Missing ("Side sheets: a genuine standard
variant"). `<x-drawer standard>` is M3's other side-sheet variant: from
`expanded` it is co-planar with the content — no scrim, no focus trap,
nothing inert, 0dp elevation, `surface` rather than surface-container-low,
no corner, and an outline-variant rule down its inner edge in place of the
scrim — sitting in the page flow at the window's full height. Below
`expanded` it is the modal sheet, since M3 caps a side sheet at 400dp and
a medium window has too little left beside one. Capped at that 400dp
whatever `width` says, and it always draws the close affordance M3
requires, because nothing else closes it there.
`pane` is untouched and the header and SKILL.md say how the two differ:
pane is the list-detail companion, standard is supplementary content.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
