orchestrate-tufte-vdqi

Name: orchestrate-tufte-vdqi
Author: gnurio

7installs18forks

How to install orchestrate-tufte-vdqi?

gnurio/tufte-vdqi-plugin/orchestrate-tufte-vdqi

$npx -y skills add gnurio/tufte-vdqi-plugin --skill orchestrate-tufte-vdqi

Installs into the current project.

Summary

Router for the Tufte data-visualization toolkit. Use whenever someone has a chart or data-visualization request and you are not sure which Tufte skill to use — it decides between assessing an existing graphic, producing a new one, or fixing a cluttered/misleading one, and chains

Files · 1

View on GitHub

SKILL.md

1	# Orchestrate Tufte VDQI
2
3	You are the router. Read the request, decide the intent, and invoke the right
4	skill. You are doing this by understanding, not by matching keywords — the
5	previous version was a brittle keyword function and it is gone.
6
7	## The toolkit (three skills + one VDQI-sourced reference)
8
9	- `assess-graphical-excellence` — evaluate an existing graphic against the nine
10	criteria, name the chartjunk species present (moiré / dreaded grid / duck /
11	decoration), compute the lie factor and compare it to VDQI's catalogue
12	(NYT MPG 14.8, TIME barrel 59.4, etc.), check whether the data deserves a
13	different genre (table for ≤20 numbers, small multiples for many series,
14	range frame instead of bordered scatter), and emit prioritised fixes tagged
15	with remedy (B1–B7), genre to switch to (C1–C10), anti-pattern resemblance,
16	and exemplar to emulate. The default when intent is unclear.
17	- `render-tufte-chart` — produce an actual chart file. Ships per-genre scripts:
18	`render_line_svg.py` (time-series C10), `small_multiples.py` (C5),
19	`quartile_plot.py` (C1), `range_frame.py` (C2, optional C3 dot-dash
20	marginals), plus `wrap_html.py` for a tufte-css HTML page. The only skill
21	that outputs a chart.
22	- `references/tufte-principles.md` (mirrored into both skills) — the canonical
23	Tufte knowledge, source-grounded to VDQI with page citations: Part A nine
24	criteria with numeric anchors, Part B seven remedies, Part C ten chart
25	genres with construction recipes, Part D chartjunk taxonomy, Part E
26	named-failure catalogue (13 dissected real-world graphics with metrics),
27	Part F named-exemplar catalogue (14 praised graphics with copyable moves),
28	Part G compact quantitative defaults.
29
30	## Routing
31
32	- Evaluate / critique ("is this chart any good?", "what's wrong with this?",
33	"is this misleading?") → `assess-graphical-excellence`.
34	- Design / build / produce ("make me a Tufte chart of…", "design a clean
35	time-series", "produce the chart") → `render-tufte-chart`. If the data is
36	currency across multiple years, deflate it first (remedy B7) before rendering.
37	- Fix / declutter an existing chart ("clean this up", "too cluttered") →
38	chain: `assess-graphical-excellence` to diagnose and list remedies, then
39	`render-tufte-chart` to rebuild honoring them. The assessment's remedy tags
40	(B1–B7) are the instructions render follows.
41	- Unsure → start with `assess-graphical-excellence`.
42
43	## Why this shape
44
45	Earlier this toolkit had ten skills. Benchmarking showed most encoded a single
46	Tufte idea the model already applies, so separate routing targets only added
47	latency and risk. Assessment and rendering are the two actions that genuinely
48	benefit from a skill. The depth that makes the skills non-generic lives in the
49	principles reference — specifically Parts C–F, the source-grounded genre
50	playbook, chartjunk taxonomy, named-failure library, and named-exemplar
51	library quoted from VDQI by page. Without those, the skills regress to
52	"Claude paraphrasing Tufte from memory"; with them, the model has Tufte's
53	specific vocabulary and worked numbers to reach for.

People who liked this also viewed

6 picks

Type

orchestrate-tufte-vdqi

7installs18forks

View on GitHub

How to install orchestrate-tufte-vdqi?

gnurio/tufte-vdqi-plugin/orchestrate-tufte-vdqi

$npx -y skills add gnurio/tufte-vdqi-plugin --skill orchestrate-tufte-vdqi

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Orchestrate Tufte VDQI
2
3	You are the router. Read the request, decide the intent, and invoke the right
4	skill. You are doing this by understanding, not by matching keywords — the
5	previous version was a brittle keyword function and it is gone.
6
7	## The toolkit (three skills + one VDQI-sourced reference)
8
9	- `assess-graphical-excellence` — evaluate an existing graphic against the nine
10	criteria, name the chartjunk species present (moiré / dreaded grid / duck /
11	decoration), compute the lie factor and compare it to VDQI's catalogue
12	(NYT MPG 14.8, TIME barrel 59.4, etc.), check whether the data deserves a
13	different genre (table for ≤20 numbers, small multiples for many series,
14	range frame instead of bordered scatter), and emit prioritised fixes tagged
15	with remedy (B1–B7), genre to switch to (C1–C10), anti-pattern resemblance,
16	and exemplar to emulate. The default when intent is unclear.
17	- `render-tufte-chart` — produce an actual chart file. Ships per-genre scripts:
18	`render_line_svg.py` (time-series C10), `small_multiples.py` (C5),
19	`quartile_plot.py` (C1), `range_frame.py` (C2, optional C3 dot-dash
20	marginals), plus `wrap_html.py` for a tufte-css HTML page. The only skill
21	that outputs a chart.
22	- `references/tufte-principles.md` (mirrored into both skills) — the canonical
23	Tufte knowledge, source-grounded to VDQI with page citations: Part A nine
24	criteria with numeric anchors, Part B seven remedies, Part C ten chart
25	genres with construction recipes, Part D chartjunk taxonomy, Part E
26	named-failure catalogue (13 dissected real-world graphics with metrics),
27	Part F named-exemplar catalogue (14 praised graphics with copyable moves),
28	Part G compact quantitative defaults.
29
30	## Routing
31
32	- Evaluate / critique ("is this chart any good?", "what's wrong with this?",
33	"is this misleading?") → `assess-graphical-excellence`.
34	- Design / build / produce ("make me a Tufte chart of…", "design a clean
35	time-series", "produce the chart") → `render-tufte-chart`. If the data is
36	currency across multiple years, deflate it first (remedy B7) before rendering.
37	- Fix / declutter an existing chart ("clean this up", "too cluttered") →
38	chain: `assess-graphical-excellence` to diagnose and list remedies, then
39	`render-tufte-chart` to rebuild honoring them. The assessment's remedy tags
40	(B1–B7) are the instructions render follows.
41	- Unsure → start with `assess-graphical-excellence`.
42
43	## Why this shape
44
45	Earlier this toolkit had ten skills. Benchmarking showed most encoded a single
46	Tufte idea the model already applies, so separate routing targets only added
47	latency and risk. Assessment and rendering are the two actions that genuinely
48	benefit from a skill. The depth that makes the skills non-generic lives in the
49	principles reference — specifically Parts C–F, the source-grounded genre
50	playbook, chartjunk taxonomy, named-failure library, and named-exemplar
51	library quoted from VDQI by page. Without those, the skills regress to
52	"Claude paraphrasing Tufte from memory"; with them, the model has Tufte's
53	specific vocabulary and worked numbers to reach for.

People who liked this also viewed

6 picks

Type