Find DataMagic recipes, adapt them to your data, and compose editable video stories.

version
0.1.0
GitHub stars
435

Install

/plugin install datamagic@datamagic

Run it inside Claude Code after adding the marketplace with /plugin marketplace add HKUSTDial/DataMagic.

Contents

Discovered from source

Skills 1

DataMagic: create with recipes

Help the user turn a chosen effect or a story goal into an editable video using the DataMagic recipe library. The gallery previews, recipe documents, template source, sample props, and this Skill form one workflow.

Choose the route

User's starting pointAction
A selected card, gallery link, slug, or copied implementation instructionsRead that recipe and adapt its source. Preserve the visual structure and reveal rhythm unless the user asks to redesign them.
Data and a goal, but no selected effectInspect the data shape, search the library, and explain suitable candidates. If asked to recommend, return recommendations; if asked to make a video, choose the best fit and proceed.
A story spanning several shotsChoose a narrative structure, match recipes to its beats, and compose them with consistent styling and timing.
An existing output to improveInspect its props, source, and render; apply the requested change and review the affected frames.

Read cards-workflow.md for recipe discovery, adaptation, rendering, and review.

Locate the recipe

Use the bundled helper from any working directory:

node <skill-directory>/scripts/cards.cjs list --query ranking
node <skill-directory>/scripts/cards.cjs inspect RankedReveal

The helper resolves the companion cards/ directory and returns absolute paths plus the render Composition ID. All 139 cards include source, schemas and sample props. Search all cards by default; --native narrows to the advanced subset when useful.

For a separately installed Skill, pass --cards /path/to/DataMagic/cards. If the companion package is missing, use an available local checkout or help obtain the requested repository before referring to its files. Honor the user's rendering environment and dependency-installation preferences.

Adapt and deliver

  • Read the chosen recipe, source, shared imports, schema, sample props and preview before adapting it.
  • Map the user's data to the actual input fields. Keep values, units, labels, ordering, display strings and claims consistent with that data.
  • Preserve the selected effect's layout and animation characteristics. Apply requested changes to text, color, media, timing or aspect ratio; portrait output needs an appropriate layout.
  • Save new props and custom source in the user's output location. Keep reusable library samples intact unless the user asks to update them.
  • Render and inspect opening, active reveal and ending frames. Check numeric accuracy, clipping, collisions and legibility. Deliver the MP4, editable files and reproducible render command.
  • Use self-review.md for the actual task; apply audio-specific checks when audio is present.

For a single template, data mapping and a short timing plan are sufficient. For a multi-shot story, read the relevant blueprint in cards/gallery/api/story-blueprints.json and scene-planning.md, then compose compatible shots. Retain each template's source clock when trimming or sequencing it.

Add deeper authoring as needed

When the user wants analysis-led authoring, a custom long-form story, an existing DVSpec edit, or a DVSpec-only plan, read full-story-workflow.md. It preserves the data-analysis, narration and DVSpec workflow for these tasks.

NeedRead
Understand columns, units or candidate findingsdata-analysis.md
Match chart form to datachart-selection.md
Organize story beatsscene-planning.md
Unify colors, typography and motiondesign-system.md
Add narration or subtitlesnarration.md, then voiceover.md
Refine or troubleshoot a resultrefinement.md, anti-patterns.md

Add voiceover when requested, align visual emphasis to measured audio, and preserve user-supplied media and provenance. If the user explicitly asks about the experimental workbench or Jianying delivery, read workbench-delivery.md; that work is currently a TODO and is separate from the default render workflow.

Sponsors

Details

Marketplace
HKUSTDial/DataMagic
Author
HKUSTDial
Category
Data Science