SLASHED — Roadmap

Current version: 0.7.30

Before v1.0

All pre-v1.0 items are complete as of v0.6.0. The API is frozen for non-additive changes; the v0.6.x series focuses on the components layer.

Post-1.0

• Components layer — the slashed.components layer is declared and its first tranche is now live: .sf-btn and .sf-card shipped in v0.7.0 (#493 / #494), including the outline .sf-btn--outline variant (#486), a button size scale, and .sf-card modifiers. The remaining .sf-* components (badge, tag, alert, avatar, modal, skeleton) stay staged in optional/components.css and land incrementally in later minors. Tracked in #384 (consolidated tranche).
• @property for spacing/shadow tokens ✓ Done — spacing (--sf-radius-*, --sf-space-*, semantic spacing) and shadow (--sf-shadow-*, --sf-text-shadow-*, --sf-drop-shadow-*) tokens are already registered via @property in core/tokens.css, alongside the existing colour registrations. Transitions on layout values and DevTools inspection both work.
• Consolidate named surfaces onto .sf-surface — the 10 .sf-surface--* variants could become one-line --sf-surface-color presets of the generic primitive, once a fallback story for the precomputed tokens.color-fallbacks.css path is settled.
• Curated core utility set — visibility/display, text-alignment, flex/grid item-alignment, gap, and content-width (default/wide/full) utilities, shipped enabled (not staged) in a new slashed.utilities layer in core/ — distinct from the staged, opt-in optional/utilities.css. Tracked in #248.

Under consideration

• Per-layer opt-in bundle via @import — already possible today via the granular core/optional files (core/layers.css first); needs docs, not code.
• Native CSS custom functions (@function) — an optional/functions.css module exposing a true --fluid(min, max) (and friends), once cross-engine support for @function exists. Until then the documented clamp() recipe covers ad-hoc fluid values without a build step.
• Heading helper utilities — activate .sf-h1–.sf-h6 in optional/utilities.css so elements can take heading typography without changing document semantics.
• Skeleton shape component — ship .sf-skeleton plus shape modifiers such as .sf-skeleton--text, .sf-skeleton--line, .sf-skeleton--avatar, and .sf-skeleton--rect in the optional components layer. Item 8 of #384. Naming decision (resolved, #575): .sf-skeleton keeps the industry-standard component name. The shimmer state was renamed .sf-is-skeleton → .sf-is-shimmer ahead of the component (shipped standalone in Unreleased — see docs/migration.md), decoupling that one breaking rename from the component work so .sf-skeleton can land purely additively. “skeleton” now means exactly one thing — the component, which will compose with .sf-is-shimmer for the shimmer effect. Shape modifiers must not reuse component names: the card-like placeholder shape is --rect (geometric), never --card, to avoid implying parity with .sf-card’s token set.
• Z-index utilities — expose the named z-index scale as .sf-z-* classes (for example .sf-z-modal and .sf-z-tooltip) in optional/utilities.css.
• First component tranche — .sf-btn and .sf-card shipped in v0.7.0 (see Post-1.0 note); .sf-badge and .sf-tag are next, using optional/tokens.components.css and existing .sf-is-* states where appropriate. Tracked in #384.
• Standalone table pattern — add .sf-table, .sf-table-scroll, and compact variants such as striped, bordered, and compact tables without changing global table styles or .sf-prose table behaviour.
• Form layout helpers — complement optional/forms.css with opt-in structure classes such as .sf-field, .sf-field__label, .sf-field__hint, .sf-field__error, .sf-form-row, .sf-form-col, and .sf-form-inline.
• Remaining optional components — after the first component tranche, complete .sf-alert, .sf-avatar, and .sf-modal with token-driven variants and native <dialog> compatibility where applicable. Items 5–7 of #384.
• Outline button variant ✓ Done (0.7.0) — the border-only .sf-btn--outline style shipped with the first component tranche. The gradient-border version shipped as the real .sf-btn--gradient modifier in 0.7.7 (fill + outline, core-4 brand families), replacing the earlier copy-paste recipe in docs/components.md. Tracked in #486 / #571.
• Decorative animation utility classes — the keyframes for sf-spin, sf-shimmer, sf-ping, sf-blink, and sf-float are already defined in core/motion.css; they need corresponding .sf-spin, .sf-shimmer, etc. utility classes in optional/utilities.css to be usable from markup.
• Hover-transform utility classes ✓ Done — .sf-hover-grow, -shrink, -float, -sink, -slide-start, -slide-end are written and documented, staged (commented out) in optional/utilities.css alongside the other opt-in utilities, pending activation. Tracked in #487.
• Overflow-detection debug utility ✓ Done — .sf-debug-overflow visually flags the element(s) causing horizontal page overflow, staged (commented out) in optional/utilities.css pending activation. Tracked in #495.
• Divide-between-children macro ✓ Done — .sf-divide (+ --vertical) applies a border between direct children in one class, reusing .sf-divider’s width/style/color tokens. Shipped (active) in core/layout.css. Tracked in #537.
• Global media radius ✓ Done — --sf-media-radius (0 by default, off) applied via :where(img, figure) in core/base.css; opt in globally by setting the token, without fighting .sf-bg-layer or a component’s own radius (layer order + zero specificity). Tracked in #529.
• Scroll-out (exit) effects ✓ Done — .sf-exit--fade/-fade-up/ -fade-down/-fade-left/-fade-right/-scale-down mirror .sf-entrance--* using animation-timeline: view() with a cover 70% → exit 100% range. Unlike entrance, animation-name is set only inside @supports (animation-timeline: view()), so engines without scroll-driven animation support leave the element visible instead of fading it out unconditionally. Shipped (active) in core/motion.css. Tracked in #527.
• Small utilities: .sf-list-none / .sf-selection--alt ✓ Done — both written and staged (commented out) in optional/utilities.css pending activation, alongside the other opt-in utilities. .sf-selection--alt overrides --sf-color-selection-bg/-text via new --sf-color-selection-*--alt tokens. Tracked in #540.
• Sticky positioning utility ✓ Done — .sf-sticky plus --s/-m/-l offset modifiers, active in optional/utilities.css, layering extra gap on top of the existing --sf-sticky-offset fluid offset. This is now the sole sticky mechanism — the redundant sticky state class was removed (position: sticky is conditional by nature and needs no runtime toggle). Tracked in #542.
• Interactive drives N exploration — turn the configurator’s passive dependency-count badge into a keyboard-accessible popover/list that navigates to dependent tokens.
• Motion configurator UX expansion — extend the existing motion domain with richer previews, motion presets, --sf-motion-scale affordances, animation tokens, and transition-token editing.
• Configurator: live hover preview in CheatsheetPanel — hovering a class/token entry in CheatsheetPanel.svelte live-previews it against the current PreviewPanel element, reverting on mouseleave. Tracked in #470.
• Design-token export formats — CSS and SLASHED theme JSON already exist; consider Figma Tokens, W3C Design Tokens, and Style Dictionary export/import if design-tool workflows become a priority. DTCG/W3C export specifically tracked in #361.
• Versioned token manifest + JSON Schema — promote docs/api-index.json to a versioned, published artifact with a companion JSON Schema so external tools (WP plugin, configurator, editors, linters) can consume the token/class catalogue against a stable contract, validated in CI. Tracked in #360.
• Consumer-facing Stylelint plugin/config — a shippable, separately published Stylelint config that lints consumer projects using SLASHED (@layer order, sf-/sf-is- reserved-name collisions, raw values that should be tokens). Tracked in #363.
• Distilled AI-assistant context file (llms.txt) — a compact, build-generated distillation of docs/llm-guide.md for consumers to drop into their own AI coding assistant’s context. Tracked in #472.
• MCP server for live token/class data — explore an MCP server exposing the generated token/class/API-index data as live tools/resources (search_tokens, get_class_info, etc.) for AI coding assistants, as a live alternative to the static llms.txt. Tracked in #473.
• Concave/inverted corner utility ✗ Not planned — shipped as .sf-corner-scoop (mask-based radial-gradient cut with 4 single-corner position variants) and then removed before 1.0. Judged too niche for the public API relative to its cost: a single absolute --sf-corner-scoop-size needs per-element tuning to read well across button-sized and hero-sized boxes, and the mask clips box-shadow/border and can’t compose with the other mask-based macros (.sf-overflow-fade, .sf-scroll-shadow) on the same element. (A two-corner “pair variant” for U-shaped cutouts had already been dropped earlier — WebKit’s mask-composite has a bug with no viable workaround.) Tracked in #484.
• Boxed-section layout primitive ✗ Not planned — a .sf-boxed section macro was implemented and then dropped after review; judged too thin a wrapper over existing primitives (.sf-section + border/shadow/ radius tokens) to justify a dedicated class. Tracked in #485.
• Flex-based grid alternative ✓ Done — shipped as .sf-grid-flex (+ size/gap/--center variants) alongside .sf-grid/.sf-equal/ .sf-grid-cols-*, documented for the uneven-item-count case. Tracked in #488.
• Unified background media overlay macro ✓ Done — resolved as composition guidance rather than a new macro: .sf-scrim composed with the existing .sf-bg-layer background-media primitive already layers a color/gradient overlay with correct automatic stacking; documented in docs/macros.md instead of adding a redundant class. Tracked in #489.
• Concrete recipe macros — partially done. Shipped: .sf-overlap/ .sf-overlap-host (overlap + card-container recipes). Two corner-recipe macros were built and then cut before 1.0 — .sf-corner-scoop (mask-based concave cut, tracked as #484 above) and a .sf-corners logical-corner recipe (with a --leaf asymmetric variant) — both judged too niche for the public API relative to their long-term maintenance cost. Tracked in #490.
• Container-relative fluid scale — the --sf-text-*/--sf-space-* fluid scales are 100vw-driven and don’t respond to @container width the way the layout primitives do; investigate a cqi-based opt-in variant for nested/narrow contexts (sidebars, cards) without replacing the viewport-based page-level default. A technique spike was posted as an issue comment; no code shipped yet. Tracked in #497.
• Hero recipe / alias — document or optionally add a .sf-hero convenience pattern that composes existing .sf-cover and .sf-scrim behaviour.
• Scroll progress bar recipe — consider an optional JS-assisted or scroll-timeline-enhanced reading progress pattern once fallback expectations are clear.
• Auto-grid ratio helper — evaluate whether a ratio-driven auto-grid adds enough value beyond the existing --sf-grid-min / .sf-grid--fit API.
• Animated link effects — extend existing link variants with opt-in effects such as underline growth or fade behaviours while preserving accessible reduced-motion fallbacks.

Out of scope

• A full utility-first API — SLASHED remains BEM-first and token-first. Small, token-backed, single-purpose helpers may be considered in optional/utilities.css, but a broad Tailwind-style utility surface is out of scope.

The following are permanently excluded — deliberate architectural decisions, not backlog. They should not be re-proposed or picked up by accident:

• Runtime JavaScript in the framework core — the core ships zero runtime JS and will keep doing so. Viewport-triggered “reveal on scroll / on-visible” behaviour that depends on an IntersectionObserver is out of scope for core/; the supported surfaces are the CSS-only .sf-visible utility hook (optional/utilities.css) and scroll-timeline view() entrance animations. Any observer-driven JS belongs to host/integration layers, never to the framework bundle.
• A consumer-facing preprocessor layer (mixins / functions) — SLASHED is authored and shipped as plain CSS with no build step for consumers. We will not expose an @include/@function-style mixin or helper-function authoring API. Native CSS (calc(), pow(), color-mix(), relative colour syntax, and — when it lands — CSS @function) is the intended substitute.
• Prefix-based automatic component styling — the framework will not auto-style arbitrary classes by name pattern (e.g. “style everything matching *--primary or .btn-*”). Components are explicit BEM classes (.sf-*); there is deliberately no “style anything that matches this prefix” mechanism.
• Viewport-breakpoint utilities / breakpoint mixins — responsiveness is expressed with container queries and intrinsic, breakpoint-free techniques. A media-query breakpoint utility or mixin system (viewport --l-/--md: suffixes, breakpoint()-style helpers) is out of scope.
• Snippet-expansion (“recipe”) syntax — a ?name-style token that expands into a CSS block requires a preprocessor or builder pass, which the pure-CSS framework will not implement. Editor/builder integrations may offer their own snippets; that is an integration concern, not a core/ feature.
• Selector-list-driven rule generation — the framework will not read a user-supplied list of selectors and graft component rules onto them (e.g. “apply these card styles to this selector list”). Consumers opt in by applying the component class itself.
• Per-channel colour partials (-h/-s/-l, -r/-g/-b) — the colour system is built on OKLCH, relative colour syntax, and color-mix(). We will not ship per-channel HSL/RGB partial tokens for manual recomposition; relative colour syntax covers those cases.
• Placeholder-content and script-wrapper helpers — generating dummy/lorem text or <script> boilerplate is not a styling concern and is out of scope for a CSS framework.