- Describe complete Fig archive, conversion, instance, and metadata ownership\n- Refresh contributor paths across localized development guides\n- Mark Track D complete in the local master plan
- Move SceneGraph-to-NodeChange conversion, text payload assembly, and blob encoding into @open-pencil/fig\n- Inject core glyph outline lookup through an explicit runtime service\n- Centralize neutral font naming in scene-graph and retain core IO/runtime orchestration
- Track normalized edited fields in SceneGraph without naming Figma payload fields\n- Preserve original raw provenance while filtering stale values through Fig-owned effective readers\n- Rewire export and renderer fallbacks and add package, SceneGraph, and exhaustive round-trip coverage
- Invalidate only edited raw Figma fields and dependent geometry or text payloads\n- Prefer untouched component and layout-grid metadata during export\n- Restore exhaustive gold fixture scene parity and refresh coverage baselines
- Add case, justification, vertical alignment, truncation, max-lines, and OpenType controls
- Apply visual text case and measured vertical offsets in CanvasKit rendering
- Preserve nullable preview values, shared text styles, font demand, and .fig metadata
- Add panel, engine, and canvas visual coverage
- Model imported fill, stroke, text, effect, and grid style definitions and references
- Add SDK and app selectors with mixed-selection batching, detach-on-edit, and undo restoration
- Preserve style definitions through .fig export and extracted subgraphs
- Document the shared-style composable and update compatibility coverage
- Add mixed-selection cap, join, and miter-limit controls with one-step undo
- Apply node miter limits across CanvasKit stroke paths and vector outline caches
- Preserve imported and edited stroke geometry through Plugin API and .fig roundtrips
- Show one variable picker affordance with a quiet identity pill and accessible combobox
- Keep bound NumberField focus non-destructive until the first value mutation
- Prefer accessible and semantic test selectors over compound test IDs
- Remove ScrubInput aliases, app wrappers, test hooks, and translated legacy references
- Share a Tailwind NumberField demo between Storybook and VitePress
- Generate component API tables from Vue metadata and type-check examples with Twoslash
- Upgrade the docs stack to VitePress 2
- Forward data-test-id attrs through rendered controls instead of exposing test-id props
- Move shell chrome components into the Shell namespace
- Remove internal planning docs from published VitePress docs
- Add guardrails and contributor guidance for test hooks and component placement
