Roadmap
Nine releases, each one usable on its own
No release depends on unbuilt future work to be valuable. Export ships before the growth engine, so the data model is validated end-to-end early and later diffing is trustworthy.
v0.1–v0.5 shipped · v0.6 in progress (Pattern Discovery & Governance Dashboard) · v0.7 and v1.0 planned
Foundation & Brand Bootstrap
shipped“Turn a logo into a real token system and a full asset kit.”
Scope
- ·Plugin skeleton, sandbox↔UI messaging, nav shell
- ·Core DesignSystemState schema + save/load round-trip
- ·Product surface multi-select + Product DNA + Tooling DNA
- ·Brand ingestion: extract fills/text/radius/shadow primitives
- ·Color Engine v1 — 6 HCT palettes, 13 tone stops, 28 roles resolved for light/dark
- ·Typography Engine v1 — 15-role Material scale, configurable modular-scale ratio
- ·Real Figma Variable Collections with alias binding
- ·Foundations page generation, idempotent
- ·Brand Kit v1 — favicon set, apple-touch-icon, manifest icons (web only)
Acceptance criteria
- ✓Bootstrap on a real brand frame produces correctly-aliased Figma Variables
- ✓Every primitive color traces to a seed hue/chroma and tone stop — no bare hex
- ✓Toggling a semantic token's mode swaps tone, not a hand-picked second value
- ✓Re-running bootstrap updates in place, never duplicates
Component Factory & Local Export
shipped“Turn Figma components into a documented, exportable library — no GitHub yet.”
Scope
- ·Component + instance scan, native property extraction
- ·Near-duplicate clustering with confirm/reject UI
- ·API contract editor — behaviorNotes, a11yNotes, codeMapping
- ·Component tokens (comp tier) bound per component/part
- ·DTCG token JSON export + foundations.md / components.md generation
- ·rules.json v1 + first CLAUDE.md / AGENTS.md scaffold (download only)
Acceptance criteria
- ✓3 near-duplicate buttons cluster and resolve correctly
- ✓Exported token JSON validates against DTCG / loads in style-dictionary
- ✓Generated CLAUDE.md contains this file's real component/token names
GitHub Sync
shipped“Design decisions leave Figma and land as a reviewable PR.”
Scope
- ·Scaffold the design-system GitHub repo structure
- ·Figma plugin Sync action → design-sync/<date>-<summary> PR with diff + impact summary
- ·CI: schema check, DTCG lint on PR; package publish on merge to main
- ·CLI v1 — design-system pull
Acceptance criteria
- ✓Editing tokens in Figma and running Sync produces a real, reviewable GitHub PR
- ✓Merging the PR triggers a package publish visible as a new version
- ✓A throwaway product repo can npm install the package and see real CSS variables
Growth Engine
shipped“The system stays accurate as designers add screens, without re-running bootstrap.”
Scope
- ·New-frame diffing against scannedFrameIds
- ·Screen Check mode — compliance scoring, issue severity tiers, jump-to-node
- ·Deterministic contrast pre-check from tone deltas
- ·Safe Auto-Fix — grouped preview, undo, nearest-tone color remaps
- ·Screen registration + automatic changelog on every mutation
- ·Drift trend across multiple scans
Acceptance criteria
- ✓A new frame with known + novel components and hardcoded colors produces an accurate diff
- ✓Safe Auto-Fix changes the compliance score correctly and is undoable
- ✓Every mutating test action has a changelog entry — zero manual logging
Developer-Side Sync
shipped“Developers building directly in code feed decisions back into the same system.”
Scope
- ·CLI scan — walk a codebase for components/hardcoded values, diff against the repo
- ·Claude Code skill + Cursor rule wrapping the same scan/propose flow
- ·dev-sync/* PR flow into the same repo as design-sync/*
- ·Code contract model — Figma variant/prop ↔ code prop mapping
- ·Prompt Seed export for Stitch / Claude chat, pre-generation
Acceptance criteria
- ✓CLI run against a sample repo with one undocumented component opens a dev-sync/* PR
- ✓That PR shows the same diff/impact shape as a design-sync/* PR
- ✓"Copy React" for a mapped component returns real import + props, not generic markup
Pattern Discovery & Governance Dashboard
in progress“Managers get a five-second read on system health.”
Scope
- ·Pattern candidate detection with confidence scoring, human approval required
- ·Maturity lifecycle actions — Promote / Reject / Merge / Keep Local / Deprecate
- ·Token gap detection (suggested, never auto-created)
- ·Governance dashboard — health score, counts, adoption %, activity, attention list
- ·Dependency graph queries + change impact analysis
Acceptance criteria
- ✓A composition repeating 3+ times across screens is proposed with correct occurrence numbers
- ✓Proposing a token change shows correct affected-screen counts before it applies
- ✓A manager with no plugin access reads system health unaided from the dashboard
Multi-Product & Brand Switching
planned“Change your logo or add a product surface, and everything downstream updates through review.”
Scope
- ·Brand Kit v2 — iOS AppIcon set, Android adaptive icons, OG card, logo lockups
- ·Bootstrap → Brand → Update — versioned, routed through design-sync/* review
- ·Multi-platform token build via Style Dictionary — CSS, JS, iOS, Android
- ·Typography Engine v2 — iOS pt / Android sp outputs from the same semantic scale
- ·Add/remove product surfaces without losing existing tokens/components
Acceptance criteria
- ✓Swapping the source logo regenerates every gated asset and tonal palette in one PR
- ✓Adding iOS after a web-only bootstrap generates iOS outputs without disturbing web ones
- ✓A manually-overridden type value on one platform survives regeneration on another
Team Governance & Versioning Polish
planned“Safe to hand to a real multi-person team.”
Scope
- ·Role mapping fully wired to GitHub permissions, role-appropriate UI gating
- ·Semantic versioning automation + structured CHANGELOG.md generation
- ·Deprecation management with migration of safe instances
- ·Design review workflow — Draft → Review → Approved → Development Ready
- ·Global design-system search + intelligent design assistant scoped to system state
Acceptance criteria
- ✓A Developer-role team member can propose via dev-sync/* but cannot merge to main
- ✓A breaking token change is classified and blocked from silent merge without major-version ack
- ✓End-to-end: logo → bootstrap → design → check → fix → register → promote → sync → ship, in one sitting
Deferred beyond v1.0 — explicit non-goals, not oversights
- ·Full multi-brand engine (distinct brands under one system) — deferred until multi-surface is proven
- ·Real-time multi-user collaborative editing inside the plugin
- ·Fully autonomous AI promotion or publishing — always human-gated, permanently
- ·Design-system marketplace / cross-org template sharing