tomevault-io Graphic Design for Diagrams
This skill takes a structural plan (which composition, which channels encode what, how elements are spatially arranged) and makes it visually excellent. It doesn't change what's encoded or which composition is used — it governs how the chosen form is rendered.
The difference between a diagram that communicates and one that communicates beautifully is entirely in this layer: the grid that creates invisible order, the type system that guides the eye, the color palette that encodes meaning without decoration, the spacing that makes grouping unambiguous.
Theoretical Foundations
Four traditions inform this skill. Understanding the why behind each rule matters more than memorizing the rule itself — the theory lets you adapt when the specific rules don't quite fit.
Gestalt Perception (Wertheimer, Koffka, Köhler)
The human visual system organizes what it sees automatically, before conscious thought. Gestalt principles describe these pre-cognitive grouping mechanisms:
Proximity — elements near each other are perceived as a group. This is the strongest grouping signal and the foundation of diagram layout. A 12pt gap between a title and its explanation says "these belong together." A 36pt gap before the next card says "new group." The viewer doesn't think about it; they just perceive the grouping. Implication: spacing IS meaning. Every gap encodes a relationship.
Similarity — elements that look alike are perceived as related. Same size, same color, same shape = same type of thing. Break similarity only to signal a different functional role — and make the break obvious (not 10.5pt vs 11pt, but 9pt vs 16pt). Implication: consistency within a type is mandatory. Inconsistency is a signal.
Closure — the mind completes incomplete shapes. A card border doesn't need to be a solid rectangle; three sides and a gap are enough. A connector doesn't need an explicit arrowhead if the layout implies direction. Implication: you can encode with less visual weight than you think, because the viewer's mind fills in the rest.
Figure-ground — the viewer separates foreground from background. Content elements (cards, labels, data marks) are figure. Background regions, subtle tints, and grid structure are ground. When ground becomes too prominent (heavy borders, saturated background fills), it competes with figure and the diagram feels cluttered. Implication: background elements should be as quiet as possible.
Continuity — the eye follows smooth paths. Aligned edges, consistent baselines, and linear connectors guide the eye smoothly. Jagged alignment and irregular spacing break the reading flow. Implication: alignment is not aesthetic preference — it's a perceptual mechanism for guiding attention.
Typography (Ellen Lupton, Thinking with Type)
Type is not decoration. It's a system of visual signals that encode hierarchy, grouping, and emphasis.
Type as hierarchy: Size, weight, and color work together to signal importance. The reader should know what's most important, second most important, and least important from the typography alone — before reading a single word.
Scale ratios: Just as musical intervals create harmony because the frequencies are mathematically related, type sizes create visual rhythm when they follow a consistent ratio. A Major Third (1.25×) or Minor Third (1.2×) ratio produces sizes that feel naturally related. Arbitrary sizes feel dissonant.
The weight system: A type family's weights (Light, Regular, Medium, Bold) are designed as a system. Three weights give you three clear hierarchy levels. Adding a fourth can work if each serves a distinct, non-adjacent function. More than four creates distinctions the eye can't reliably make at body sizes.
Spacing as content: The space between letters (tracking), between lines (leading), and between text blocks (margins) carries meaning. Tight tracking on uppercase text crowds the letterforms. Generous leading on multi-line text prevents line-skipping. These aren't preferences — they're functional requirements with perceptual thresholds.
Grid Systems (Josef Müller-Brockmann)
The grid is the invisible infrastructure that creates visual order. Viewers never see the grid, but they perceive its presence as coherence and its absence as chaos.
The base unit: Every spatial measurement in a diagram — margins, gutters, padding, element dimensions — should be a multiple of a single base unit (typically 8pt or 12pt). This constraint produces visual harmony because all proportions are related. Off-grid placement by a few points here and there registers as sloppiness even though nobody can articulate why.
Margins and gutters: Margins (page-edge breathing room) should be at least as generous as gutters (inter-element spacing). When gutters exceed margins, content feels like it's spilling off the page.
Content density: Content should fill 70-85% of the available area. Below 70% feels empty. Above 85% feels cramped. For fixed page sizes, work backward from dimensions to determine the maximum content rectangle.
The grid serves the content, not the reverse. When content doesn't fit the grid cleanly (an odd number of cards, a diagram that's wider than it is tall), adjust the grid — don't distort the content. Müller-Brockmann's grids are starting points, not prisons.
Color Theory (Josef Albers, Johannes Itten)
Color in information design is functional, not decorative. Every color decision should answer: "What information does this color communicate?"
Albers' interaction of color: Colors don't exist in isolation — they change in the presence of their neighbors. A gray square on a white background looks darker than the same gray on a black background. Implication: test colors in context, not in swatches.
Value (lightness) before hue: In Itten's framework, value does more work than hue. Dark-on-light creates hierarchy. Light text on dark backgrounds draws attention. Value contrast is what makes text readable (WCAG 4.5:1 minimum). Hue is for categorization; value is for hierarchy.
The 60-30-10 rule: 60% dominant (white/near-white background), 30% secondary (grays for text, borders, structural elements), 10% accent (one saturated hue for emphasis). When accent exceeds ~10%, it stops functioning as emphasis — it becomes just another color. Restraint is what makes the accent powerful.
Simultaneous contrast: A blue accent strip next to a gray card will make the gray appear slightly warm (yellowish). A warm orange next to neutral gray will make the gray appear cool (bluish). This matters when grays are used for hierarchy — the "same gray" can look different depending on what's next to it. Verify by checking colors in their actual layout context.
Core Workflow
Phase 1: Establish the Grid
Given the composition and element count from Skill 2, design the grid system.
Step 1: Choose the base unit. 8pt for compact diagrams (many elements, screen display). 12pt for spacious diagrams (fewer elements, print, presentations). The base unit determines everything else.
Step 2: Define margins and gutters.
Margins (page edge): 3-4 × base unit minimum
Gutters (between elements): 2-3 × base unit
Internal padding (within elements): 1-2 × base unit
Rule: margins ≥ gutters ≥ internal padding. This creates three spatial levels that encode containment hierarchy.
Step 3: Calculate element dimensions. Work from the page/artboard size inward:
Available width = artboard width - 2 × margin
Element width = (available width - (cols-1) × gutter) / cols
Element height = derived from content needs (text lines, internal spacing)
Round element dimensions to the nearest base unit.
Step 4: Verify content density. Calculate total content area / artboard area. Should be 0.70-0.85. If too sparse, reduce margins or increase element size. If too dense, add margins or reduce content (go back to Skill 1 and cut more).
Step 5: Check Gestalt proximity ratios. Intra-group spacing (within a card or node) should be ≤ half of inter-group spacing (between cards or nodes). This ensures proximity-based grouping reads correctly.
Output: A grid specification with base unit, margins, gutters, element dimensions, all as multiples of the base unit.
Phase 2: Design the Type System
Step 1: Choose the type family. For diagrams, sans-serif families with clear weight differentiation work best. Helvetica Neue, Inter, and system sans-serifs are reliable. One family only — introducing a second family is rarely justified and creates cognitive overhead.
Step 2: Define the scale. Pick a ratio and derive all sizes from the base:
Ratio: 1.22× (Minor Third) — good for compact diagrams
Ratio: 1.25× (Major Third) — good for spacious diagrams
Ratio: 1.33× (Perfect Fourth) — good for presentations
Example (base 9pt, ratio 1.22×):
9pt body text, annotations, levers
11pt card titles, section labels (9 × 1.22)
13pt (usually skipped — reserved for emphasis if needed)
18pt page title (9 × 2, or approximately 9 × 1.22³)
Every text element must use a size from this scale. No exceptions.
Step 3: Assign weights to hierarchy levels.
Level 1: [largest size] + Bold → page title, entry point
Level 2: [medium size] + Bold → section/card titles
Level 3: [base size] + Regular → body text, explanations
Level 3b: [base size] + Medium → insights, emphasis (accent color)
Level 4: [base size] + Regular → secondary text (mid-gray)
Maximum 3-4 levels. Each level must be visually distinct from adjacent levels at their rendered size. If two levels look the same when rendered, merge them.
Step 4: Set spacing.
- Leading (line spacing): ≥ 1.3× font size. For 9pt text, ≥ 12pt leading.
- Tracking for ALL-CAPS: +50 to +120 (uppercase letterforms need room)
- Tracking for body text: 0 (default)
- Line length for multi-line text: 45-75 characters (optimal: 66)
Output: Type specification with family, scale ratio, sizes, weights, leading, tracking, and the hierarchy level each combination serves.
Phase 3: Design the Color Palette
Step 1: Determine functional color roles. Every color in the diagram must have a job:
Background: white or near-white (rgb 255,255,255 or 249,250,251)
Primary text: dark gray (rgb 31,41,55) — contrast ≥14:1 on white
Secondary text: mid-gray (rgb 107,114,128) — contrast ≥4.5:1 on white
Accent: one saturated hue — contrast ≥4.5:1 on white
Structure: light gray for borders, rules, gridlines
Grouping tint: very light tint for background regions (if needed)
Two text grays only. Primary dark for text that matters. Mid-gray for secondary text. A third gray creates distinctions the eye can't reliably perceive at body sizes.
Step 2: Choose the accent hue. The accent does double duty: it signals emphasis AND encodes one categorical dimension (if Skill 2 assigned color hue to a dimension). If color hue encodes a category, the accent hue is that encoding. If not, choose a hue that contrasts with the gray palette without being aggressive. Blues and teals are safe defaults.
Step 3: Verify contrast ratios. WCAG AA minimum:
Normal text (< 14pt bold, < 18pt regular): ≥ 4.5:1
Large text (≥ 14pt bold, ≥ 18pt regular): ≥ 3:1
Check every text color against its background. Common failure: mid-gray text on a tinted background drops below 4.5:1. Common fix: darken the text or lighten the background.
Step 4: Check the 60-30-10 distribution. Visually estimate: is the accent (saturated hue) ≤ 10% of visual area? If it appears on backgrounds, borders, or body text, it's overused and no longer functions as emphasis.
Step 5: Verify grayscale readability. Convert the palette to grayscale. If any two elements that should be distinguishable become indistinguishable, add a non-color differentiator (position, shape, weight, or text label).
Output: Color palette with RGB values, functional role for each color, contrast ratios verified, 60-30-10 distribution confirmed.
Phase 4: Define Element Styling
Step 1: Stroke weight hierarchy. Maximum 4 levels, each serving a distinct function:
0.25-0.5pt structural rules (separators, gridlines)
0.5-1pt containment borders (card outlines, region boundaries)
1-2pt relational connectors (arrows, flow lines)
2-3pt emphasis elements (accent lines, pull-quote bars)
Step 2: Corner radius. If using rounded rectangles, one radius value for all elements. Typically 4-6pt. Consistency matters more than the specific radius.
Step 3: Label positioning rules. Labels within 4pt of their parent element. Arrow labels on the arrow line, not floating in gutters. The Gestalt proximity principle: if a label is equidistant from two elements, it belongs to neither.
Step 4: Consistency audit. Same-type elements (cards, nodes, connectors) must have identical treatment — same dimensions, same stroke weight, same internal spacing, same type treatment. Break similarity only to signal a different functional role, and make the break obvious (change 2+ attributes, not just one).
Output: Element styling spec with stroke weights, corner radii, label rules, and consistency requirements.
Phase 5: Compose and Audit
Step 1: Assemble the full specification. Combine grid, type, color, and element styling into a complete design spec.
Step 2: Check hierarchy alignment. Size, weight, color, and position should all agree about what's most important. If the largest text is at the bottom, the boldest text is a subtitle, and the accent color highlights a footnote, the hierarchy signals conflict and the viewer's eye bounces.
Step 3: Entry point test. Squint at the planned layout (or imagine it at 25% zoom). Exactly one element should dominate. If two compete, strengthen the winner or weaken the competitor.
Step 4: Tufte's data-ink audit (inherited from Skill 2). Review every planned visual element: does it encode data, provide necessary structure, or earn its place through grouping/hierarchy? If not, remove it.
Step 5: Symmetry decision. Per section of the diagram, choose symmetrical (equal treatment = peer status) or asymmetrical (unequal treatment = hierarchy) balance deliberately. Don't mix within a section.
Read references/design-rules.md for the full 24-rule audit checklist.
Phase 6: Produce the Design Specification
Output the complete specification that the rendering skill consumes.
Structured specification (YAML):
grid:
base_unit: 12
margins: { left: 48, right: 48, top: 48, bottom: 48 }
gutters: { horizontal: 36, vertical: 24 }
internal_padding: { horizontal: 12, vertical: 8 }
artboard: { width: 612, height: 792 } # letter = 612×792pt
content_density: 0.78
type:
family: "HelveticaNeue"
scale_ratio: 1.22
levels:
- level: 1
size: 18
weight: Bold
tracking: 0
role: "page title"
- level: 2
size: 11
weight: Bold
tracking: 0
role: "card/section titles"
- level: 3
size: 9
weight: Regular
tracking: 0
role: "body text, explanations"
- level: 3b
size: 9
weight: Medium
color_override: accent
role: "insights, emphasis"
- level: 4
size: 9
weight: Regular
color_override: secondary
role: "levers, annotations, footer"
leading_ratio: 1.33
uppercase_tracking: 80
max_line_characters: 65
color:
background: [255, 255, 255]
primary_text: [31, 41, 55]
secondary_text: [107, 114, 128]
accent: [37, 99, 235]
structure_border: [209, 213, 219]
structure_rule: [229, 231, 235]
grouping_tint: [243, 244, 246]
contrast_verified: true
grayscale_verified: true
elements:
stroke_weights:
structural: 0.5
containment: 0.75
connectors: 1.5
emphasis: 2.5
corner_radius: 4
label_max_distance: 4 # pt from parent element
accent_strip:
concept_width: 2
synthesis_width: 4
composition_rules:
hierarchy_levels: 4
entry_point: "title, top-left"
reading_path: "title → first card → arrow → second card → ..."
balance: "symmetrical for concept grid, asymmetrical for synthesis card"
data_ink_notes: "no decorative borders, no non-functional tints"
References
references/design-rules.md— The full 24-rule audit checklist organized by Grid & Layout, Typography, Color, Visual Elements, and Composition. Each rule includes the principle, the perceptual reason, and how to verify compliance. Run through this after every render.
Converted and distributed by TomeVault — claim your Tome and manage your conversions.