Dataviz Kit

A spec-driven toolkit for building executive dashboards and consulting storytelling that respect the fundamentals of data visualization — Tufte, Shneiderman, Ben Fry, Munzner, Few, and Minto.

Two parts, one workflow:

• ux-spec-skills/ — a reusable Spec Skills Library. Skills turn a request ("I need a weekly exec review") into a structured spec; templates give you doc skeletons; examples show finished outputs; a critique skill reviews any dashboard/chart against the principles.
• pipeline/ — a dataset-agnostic CSV → spec → self-contained HTML dashboard pipeline. Point it at your CSVs and get a profile, a spec, and a ready-to-open dashboard.

Why

Most dashboards are built by taste and habit. This kit encodes the principles into repeatable skills and a small pipeline, so every spec and output follows the same non-negotiables:

• Tufte — clarity and integrity; signal over noise; no chartjunk.
• Shneiderman — overview first, then zoom/filter, then details on demand.
• Ben Fry — Acquire → Parse → Filter → Mine → Represent → Refine → Interact.

The workflow

Inputs (domain, KPIs, cadence, audience)
    → Skill (e.g. daily-exec-pulse, weekly-performance-review)
    → Spec (purpose, modules, data, interaction, trust, states, events)
    → Dashboard build (design + engineering)
    → Optional: consulting outputs (exec slide outline, one-pager)

Quick start

Generate a spec (no install needed):

1. Open ux-spec-skills/SKILLS_INDEX.md (or the router at ux-spec-skills/skills/99-router.skill.md) and pick a skill.
2. Fill in the skill's "Inputs" section and run its "Prompt" in Cursor (or your tool of choice).
3. Optionally start from a matching template in ux-spec-skills/templates/.

Turn CSVs into a dashboard:

pip install -r pipeline/requirements.txt
cp datasets/dashboard.example.yaml datasets/dashboard.yaml   # edit paths to your CSVs
python3 pipeline/run.py                                       # → output/<name>.dashboard.html

See pipeline/README.md for details.

Repo layout

Path	Purpose
ux-spec-skills/	Spec Skills Library: skills, templates, examples, expert demos, principle references. Start at its README.md.
pipeline/	CSV → spec → self-contained HTML pipeline (Python).
datasets/	Pipeline config. dashboard.example.yaml is committed; your dashboard.yaml is local/gitignored.
data/	Your local CSVs. Gitignored so PII never enters the repo (only data/README.md is committed).
output/	Generated artifacts (gitignored).
landing/	Static landing page for the kit (see landing/README.md).

Principles

Detailed summaries live in ux-spec-skills/reference/: Tufte, Shneiderman, Ben Fry, Stephen Few, Tamara Munzner, Barbara Minto, and Storytelling with Data.

License

MIT — fork it, make it yours.