Skip to main content
skill CX Codex

design-system-importer — Claude Code skill: design references

This skill turns reference media and brand materials into visual constraints for a video project. Use it to reproduce a style or adapt selected parts of it.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add Orkas-AI/Orkas-VideoStudio --skill design-system-importer -a claude-code
Open on GitHub

What design-system-importer does

The skill reads images, videos, screenshots, DESIGN.md, brand guides, and design notes. It first classifies the requested relationship to each reference: reproduce it, edit the original, or borrow selected principles. Then it records style tokens and defines what must stay intact and what can change.

The result feeds project/composition/composition-manifest.json#art_direction and guides hand-authored HTML/CSS/SVG. For web and video references, it accounts for composition, timing, safe zones, and motion; subtitles remain in tracks.captions.lines. Inspected material is copied to project/composition/assets/references/. With only vague adjectives and no named source, the skill directs aesthetic decisions to frontend-design instead.

What it helps with

  • carry a brand guide’s visual rules into a video
  • define which parts of a reference must remain unchanged
  • map motion from a video reference to project scenes
  • adapt a website’s visual language for video

Example prompts for the agent

  • Use DESIGN.md to extract the visual rules for this video project.
  • Treat this screenshot as a reference: preserve its composition and change the palette.
  • Map the motion in this video reference to the project’s target scenes.

How to install design-system-importer

1Skills CLI (Vercel)

npx skills add Orkas-AI/Orkas-VideoStudio --skill design-system-importer -a claude-code

2Manually: put the files in place

~/.claude/skills/design-system-importer/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/design-system-importer/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about design-system-importer

How do I install design-system-importer in Claude Code?

Run npx skills add Orkas-AI/Orkas-VideoStudio --skill design-system-importer -a claude-code. The skill file is placed at ~/.claude/skills/design-system-importer/SKILL.md.

What does design-system-importer do?

It extracts concise style tokens from references and records which visual elements to preserve or allow to change in a video project.

When should I use this skill?

Use it when a task has a concrete style source, such as a screenshot or brand guide. For ordinary editing, TTS, or clip selection, the skill says not to use it; vague style requests without a source go to frontend-design.

Source text

SKILL.md · MIT
---
name: design-system-importer
description: Reference-media and design-system input layer for OrkasVideoStudio. Treat reference images, videos, screenshots, brand guides, and design notes uniformly, compiling reproduce/edit/guide intent into executable spatial and temporal constraints.
---

# design-system-importer

Use this when COMPOSE or an AUTO compose segment has external reference media or a style source: an image, video, `DESIGN.md`, brand guide, screenshot, existing website, design notes, or an explicit named visual direction.

Do not use it for ordinary editing, TTS, shot generation, or clip selection. Do not introduce a new user Gate. The output is an internal style extraction that feeds `project/composition/composition-manifest.json#art_direction` and the hand-authored `project/composition/index.html`.

Do not use it for vague adjectives like "modern", "clean", "premium", "dynamic", or "more polished" when no source is named. In those cases, let `frontend-design` choose the aesthetic thesis directly from the video brief.

## Reference Intent Before Input Technique

For every supplied image or video, classify the requested relationship before extracting style:

- `reproduce`: preserve the declared content, identity, composition, structure, style, motion, timing, or audio axes.
- `edit`: use the media as the original, protect unaffected axes, and change only `may_change`.
- `guide`: borrow only declared roles without implying exact fidelity. This is the safe default when intent is unspecified.

Use `intent_basis:"user"` for explicit requirements and `"inferred"` only for a fallback. The contract depends on requested intent, not whether the pixels came from a camera, website, design tool, model, or another authoring format. Copy each inspected source into `project/composition/assets/references/` and reference that stable local path.

Follow the requested `reproduce`/`edit`/`guide` relationship. Preserve user-owned or explicitly authorized assets when fidelity requires them; otherwise respect stated exclusions and ownership constraints. Do not force variation merely to appear original when the user asked to reproduce a source.
Keep extraction small enough to fit inside the design contract. Do not load or recreate unrelated parts of an external design system.

## Extract Compact Tokens

Write a `style_source` object into `project/composition/composition-manifest.json#art_direction`:

```json
{
  "style_source": {
    "source_type": "brand_system | design_notes | reference_media | existing_product | named_reference",
    "source_basis": "file path, user note, or inspected artifact",
    "adaptation_boundary": "what must be preserved, may change, or is excluded",
    "observed_signature": "the concrete geometry, hierarchy, type, palette, and motion behavior seen in evidence",
    "confidence": "high | medium | low",
    "fidelity_mode": "exact | close | adapt"
  }
}
```

Then normalize the source into tokens that hand-authored HTML/CSS/SVG can consume:

- `color_tokens`: background, surface, text, muted, primary accent, optional secondary accent, plus intended contrast relationship.
- `typography_tokens`: display, body, data/label, caption roles; scale and weight intent; avoid relying on fonts that are not available.
- `shape_tokens`: radius, stroke, shadow, divider, border, and density.
- `layout_language`: grid, editorial, cinematic, dashboard, diagrammatic, poster, product-demo, or another concrete grammar.
- `motion_language`: entrance, transition, emphasis, data-build, and exit patterns; keep it compatible with GSAP timeline seeking.
- `asset_rules`: what images/icons/marks are allowed, need replacement, or must be avoided.
- `excluded_elements`: user exclusions and any assets that cannot be reused; do not invent blanket exclusions that contradict an explicit reproduce/edit request.

Keep the imported style small. If more than 6 chromatic colors or 3 font roles are needed, summarize the conflict and pick the smallest faithful subset.

## Executable Media Contract

For every concrete reference, add `art_direction.references` with `id`, `media_type`, local `path`, `intent`, `intent_basis`, allowed `roles`, `required`, `preserve`, `may_change`, and `target_scene_ids`. Use only these roles: `content`, `identity`, `composition`, `structure`, `style`, `motion`, `timing`, and `audio`.

Add shared `art_direction.reference_fidelity` with `mode: exact|close|adapt`, non-overlapping `preserve`/`may_change`, normalized `layout_anchors` for composition/structure roles, and concrete observable comparison criteria. Video reproduce/edit/motion/timing references also need source-time-to-target-scene `temporal_anchors`.

`exact` preserves every declared axis except explicit `may_change`; `close` keeps the recognizable system while adapting named content or aspect constraints; `adapt` borrows only selected declared principles. Review with paired source/target frames and concrete findings, never a numeric similarity score.

## Map To Video

Web and brand systems are not videos. Convert them for motion:

- First frame: choose the style's strongest thumbnail-friendly signal.
- Safe zones: enlarge type and spacing beyond web density.
- Scene variation: turn repeated web sections into distinct beats.
- Motion: make the brand grammar move with purpose; do not animate every component.
- Captions: keep ordinary subtitles in `tracks.captions.lines`, not in the style system.

## Output

After extraction, the design contract must state:

- What source was used.
- Which tokens were adopted.
- Which tokens were deliberately simplified.
- Which elements are excluded and why.
- What observed visual signature must remain recognizable for the declared intent.
- Which reference intent, roles, protected/allowed changes, target scenes, anchors, and observable comparison criteria apply.

© Orkas-AI · MIT · Orkas-AI/Orkas-VideoStudio

Similar skills

Everything for Design →

From the same set14

stage-decide Inspects raw footage to remove dead air, cut filler words, and rank repeated takes into an evidence-based rough cut. Useful when turning long recordings into tight edits. CX Codex +1 ★ 329repo stars stage-consistency Keeps characters, framing, and visual style aligned across shots and scenes. It targets multi-shot videos and long stories, not one-off generation. CX Codex +1 ★ 329repo stars stage-compose This skill gives the agent authoring knowledge for Orkas/OVS HTML compositions and runs them through ovs draft to produce an mp4. Useful for explainers, titles and motion graphics. CX Codex +1 ★ 329repo stars stage-assemble Assembles a validated cross-modal EDL from plan.json into a final video using ovs CLI and FFmpeg tiers. Runs only after plan approval. CX Codex +1 ★ 329repo stars orchestration Coordinates video creation and editing from the initial concept to the final export, pausing for user approval before important production decisions. CX Codex +1 ★ 329repo stars gate-control Sets the rules for what a video approval actually authorizes and which step comes next. Useful once edits start arriving after a gate. CX Codex +1 ★ 329repo stars frontend-design Directs aesthetics for OrkasVideoStudio HTML and SVG motion graphics before writing index.html. It prevents generic neon layouts and unreadable video type. CX Codex +1 ★ 329repo stars composition-design-review A design review layer for OrkasVideoStudio COMPOSE previews and drafts. It reads the whole frame index, then opens the cover and risky frames at full size before any preview is shown. CX Codex +1 ★ 329repo stars

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor