DreambaseRegistry
BlocksSkills

Getting started

  • Introduction

Blocks

  • Overview
  • Questionnaire CardNew

Skills

  • Overview
  • data-presentation
  • data-stories
  • echarts
  • industrial-schematics
  • mcp
  • micrographic-design
  • public-reports
  • skill-creator
  • visualization-design

Built by Dreambase. Open source under MIT.

GitHub

Skill

dreambase-data-stories

Design best-in-class data-story artifacts — reports, executive summaries, slide decks, one-pagers, infographics, and scrollytelling pieces — that blend accurate charts, illustration, great typography, and narrative structure into easy-to-consume stories. Use whenever the user wants to present, summarize, or communicate analytics or data to an audience: "make a report", "exec summary", "board deck", "slides", "one-pager", "data story", "writeup of these results", or "make this presentable" — even if they don't mention design. Always apply dreambase-visualization-design for the charts themselves; use dreambase-echarts when charts will actually render.

evalsreferences
npx skills add DreambaseAI/skills --skill dreambase-data-stories
View on GitHub

Data Stories

Turn analysis into artifacts people actually read, understand, and act on. The bar is the work of the field's best — FT, NYT Graphics, The Pudding, Reuters — where accurate charts, deliberate typography, and narrative structure combine into something that reads effortlessly. "Mechanically correct content in default styling" is the failure mode this skill exists to prevent.

Skill composition (non-optional wiring)

  • dreambase-visualization-design — always. Every chart in every artifact follows its selection matrix, encoding rules, and integrity gate. This skill adds the narrative and layout layers on top; it never overrides honest scales, context, or accessibility. Beauty doesn't buy exemptions.
  • dreambase-echarts — when charts render. Standard forms from its references; bespoke/illustrative forms via its custom-series reference (charts/custom.md in that skill — custom series, pictorialBar, graphic layer).
  • Mechanics skills — when producing files. If available, delegate .pptx to the pptx skill, .docx to docx, PDF to pdf; this skill owns the design system those mechanics execute.

Core stance

  1. Story before design, design before build. The Big Idea (one complete sentence: point of view + stakes) exists before any layout; if it can't be written, return to the analysis.
  2. Answer first for deciders, arc for readers. Executive artifacts use the pyramid: conclusion, then support. Narrative artifacts use hook → context → build → insight → resolution. Know which you're making.
  3. Annotation is narration. The text layer — claim-titles, direct labels, callouts — is the product, not garnish. Edit chart text as rigorously as body copy.
  4. One idea per view for scrolling and deciding audiences; density only for audiences who will linger.
  5. Scroll over click. Nothing load-bearing hides in a tooltip or behind interaction; clicks must deliver something spectacular.
  6. Bespoke when the story demands it, standard by default. Custom forms make the exceptional story memorable; automation and standard forms carry the daily work.
  7. Beauty earns attention; encoding earns trust. Spend craft on typography, composition, and the annotation layer — never on distorting marks.
  8. Design for the constrained reader: mobile, print, screen-reader, forwarded-without-context.

These distill the verified practice of Bremer, Lupi, The Pudding, NYT Graphics, Bostock, Burn-Murdoch, Reuters Graphics, McCandless (and his critics), and FiveThirtyEight — including where they disagree. Read references/storytellers.md for the full principles, quotes, and contested points.

Workflow

  1. Frame — audience, decision or takeaway, format, tier (exec/manager/analyst), constraints. Write the Big Idea sentence. (references/narrative-structure.md)
  2. Outline the argument — pyramid for exec artifacts, arc for narrative ones. Slide titles / section heads must read as a coherent argument alone.
  3. Choose the evidence — for each point, the one chart or number that proves it. Run every chart decision through dreambase-visualization-design (its matrix, then its integrity gate). Identify the hero chart — the one that is the argument.
  4. Design the system — type pairing and scale, 4–6 color palette, grid — once, for all artifacts from this analysis. (references/page-design.md)
  5. Integrate — charts share the artifact's type and palette; annotation carries the narration; illustration frames but never encodes. (references/chart-integration.md)
  6. Build — via the mechanics skills or clean HTML; bespoke chart forms via dreambase-echarts custom toolkit.
  7. Test the cold reader — can someone with zero context extract the takeaway (5 seconds for an infographic, 30 seconds for an exec summary)? Fix the artifact, not the reader.

References — read what the task needs

  • references/narrative-structure.md — pyramid principle/BLUF, the Big Idea, story arcs, exec-summary/slide/report/infographic conventions, presentation-vs-slidedoc, audience tiering. Read first for any new artifact.
  • references/page-design.md — typography system, numeric craft (tabular figures, real minus signs), grid, color-as-system, tables, big-number callouts, format-mechanics delegation. Read when designing the artifact's look.
  • references/chart-integration.md — hero charts, assertion-evidence titling, annotation-as-narration, making charts belong to the design, illustration rules, interactivity budget, the pre-ship checklist. Read when placing charts into layouts.
  • references/storytellers.md — the verified practice of the field's masters, the 12 cross-cutting principles, and which are contested. Read when choosing a storytelling approach or justifying a bespoke form.

Output discipline

Deliverables state their format, audience tier, and Big Idea up front, and always include the design layer — hierarchy, callouts, type and color intent — even when the output format is plain markdown; the content alone is half the deliverable. Every chart passes the visualization-design integrity gate; every artifact passes the cold-reader test; observation, interpretation, and recommendation stay visibly separated. When asked to make data "more impressive" than it is, apply the integrity playbook: explain the risk, deliver what the data honestly supports, styled so well nobody misses the distortion.