graphic-universe-builder

Builds a project-specific graphic universe by researching visual territories beyond conventional web-design references, clustering recurring forms and ideas, identifying category clichés, and deriving

SKILL.md

GitHub

Graphic Universe Builder

Purpose

Build a project-specific visual research foundation before interface design begins.

The skill expands research beyond current website trends and competitor screenshots. It studies the project's business, language, objects, processes, history, environment, materials, culture, adjacent disciplines, and abstract representations to uncover recurring visual patterns.

Those patterns are then synthesized into visual vocabulary, metaphor candidates, and several candidate graphic universes that can inform later creative-direction work.

The goal is not to collect attractive references.

The goal is to discover a visual world that could plausibly belong to this project and not five unrelated competitors.

This skill performs research and synthesis. It does not design the final interface.


Core principle

Trace important visual decisions back to something meaningful about the project.

Prefer:

project evidence
→ research territory
→ observed pattern
→ intended meaning
→ possible visual application

Avoid:

current design trend
→ attractive reference
→ copied aesthetic

A strong graphic universe gives later design decisions a reason to exist.


When to use this skill

Use this skill when:

  • Starting the visual direction for a new website, landing page, campaign, portfolio, editorial experience, or digital brand.
  • The user asks to build a graphic universe, visual world, moodboard direction, reference system, or visual research foundation.
  • A project brief exists but the visual language has not yet been determined.
  • An existing design feels generic, trend-led, overly SaaS-like, or interchangeable with competitors.
  • The team keeps using the same visual references across unrelated clients.
  • The user wants visual inspiration without simply copying websites from Behance, Dribbble, Awwwards, Pinterest, or similar platforms.
  • A project needs meaningful visual metaphors tied to its product, service, history, process, audience, or subject.
  • Several creative directions will later be explored and they need stronger source material first.
  • A seemingly visually uninteresting industry needs a richer source of creative material.
  • The designer wants to understand what visual ingredients are actually specific to the project.

Do not use this skill when:

  • The user already has several mature visual directions and only needs to choose between them.
  • The task is primarily UI usability, information architecture, accessibility remediation, conversion optimization, or implementation.
  • The user only wants a quick collection of attractive websites.
  • The user wants final page mockups, production UI, CSS, components, or code.
  • A complete visual system already exists and the task is only to apply it consistently.
  • The problem is still fundamentally unclear and the business, audience, product, or communication objective has not been sufficiently defined.

If the project itself is poorly understood, run a design-problem-framing workflow before deep visual research.


Inputs to inspect

Inspect as many of the following as are available:

Project context

  • Project brief
  • Business goals
  • Product or service description
  • Audience or customer segments
  • Desired user actions
  • Key messages
  • Brand positioning
  • Existing strategy documents

Brand material

  • Logo
  • Brand guidelines
  • Existing typography
  • Color system
  • Photography
  • Illustration
  • Packaging
  • Printed material
  • Advertising
  • Archive material
  • Previous websites

Product and operational material

  • Physical products
  • Materials
  • Manufacturing processes
  • Tools
  • Workflows
  • Interfaces
  • Locations
  • Facilities
  • Environments
  • Machinery
  • Diagrams
  • Maps
  • Documentation
  • Data structures

Narrative material

  • Founder story
  • Company history
  • Customer stories
  • Product origin
  • Naming
  • Terminology
  • Internal language
  • Cultural context
  • Geographic context
  • Important events or milestones

Existing visual references

  • User-provided references
  • Moodboards
  • Competitors
  • Screenshots
  • Photography
  • Art direction examples
  • Editorial references
  • Previous concept work

When external research is possible, expand beyond the materials supplied by the user.


Research mindset

Do not begin by asking:

What websites look good for this industry?

Begin by asking:

What visual world naturally surrounds this subject?

Research both close and distant relationships.

Move progressively from the obvious toward less commonly referenced sources.

Useful sources can include:

  • Product and industrial design
  • Architecture
  • Editorial design
  • Books
  • Historical publications
  • Packaging
  • Signage
  • Exhibition design
  • Photography
  • Film
  • Title sequences
  • Fine art
  • Illustration
  • Technical drawings
  • Diagrams
  • Maps
  • Patents
  • Machinery
  • Manufacturing
  • Materials
  • Fashion
  • Cultural artifacts
  • Scientific imagery
  • Natural structures and systems

Current website references may be included, but they should not dominate the research.


Workflow

1. Establish the visual research question

Before collecting references, summarize:

  • What the project is.
  • Who it is for.
  • What makes it different.
  • What it should communicate.
  • What emotional impression may be appropriate.
  • What currently makes the category visually repetitive.
  • What the research needs to discover.

Create a short statement:

We are researching how <project-specific idea> could become a distinctive
visual language without relying on <obvious category cliché or trend>.

Do not silently invent missing business strategy.

Mark uncertain assumptions explicitly.


2. Build the project vocabulary

Extract words from the project material before searching visually.

Organize them into these groups:

Group Look for
Nouns Products, objects, places, people, tools, components
Verbs Actions, processes, transformations, movements
Materials Metal, paper, glass, fabric, code, light, soil, data
Forms Circles, joints, folds, grids, waves, layers, fragments
Environments Factory, coast, laboratory, studio, street, archive
Qualities Precise, raw, quiet, dense, playful, resilient
Emotions Trust, anticipation, curiosity, freedom, tension
History Origins, milestones, eras, cultural references
Language Repeated words, names, jargon, phrases, idioms
Relationships Connection, separation, flow, exchange, hierarchy
Transformation Before/after states created by the product or service

Prioritize vocabulary that is specific to this project rather than generic branding adjectives.

Avoid relying on words such as:

  • modern
  • clean
  • innovative
  • premium
  • bold
  • minimal
  • futuristic

unless the project provides a concrete interpretation of them.


3. Create the research territory map

Turn the vocabulary into distinct research territories.

At minimum consider:

A. Direct territory

The project's category, product, service, and immediate environment.

Examples:

furniture
metal fabrication
wedding photography
robotics
financial reporting
music production

B. Object and process territory

Investigate how things are constructed, transformed, used, repaired, transported, or operated.

Look for:

  • Tools
  • Parts
  • Joints
  • Fasteners
  • Machinery
  • Materials
  • Stages
  • Movements
  • Physical transformations

C. Abstract representation territory

Explore how the subject has been reduced into symbols or systems.

Look for:

  • Icons
  • Diagrams
  • Maps
  • Schematics
  • Symbols
  • Marks
  • Logos
  • Notation
  • Charts
  • Geometry
  • Patterns

D. Brand and historical territory

Investigate:

  • Founder history
  • Archive imagery
  • Historical products
  • Previous identities
  • Old advertising
  • Geographic history
  • Cultural context
  • Significant eras

E. Adjacent discipline territory

Move outside interface design.

Possible territories include:

  • Architecture
  • Industrial design
  • Editorial design
  • Packaging
  • Fashion
  • Film
  • Photography
  • Exhibition design
  • Signage
  • Print
  • Fine art

F. Structural and natural territory

Explore deeper formal analogies.

Possible sources include:

  • Geological formations
  • Cellular structures
  • Networks
  • Branching systems
  • Erosion
  • Waves
  • Growth
  • Repetition
  • Symmetry
  • Fracture
  • Layering
  • Light
  • Shadow

Use these only when the connection is meaningful.

Do not add nature references merely because they look interesting.

G. Opposite territory

Research the visual opposite of the category convention.

Examples:

software → physical machinery
luxury → utilitarian manuals
industrial → delicate botanical structures
finance → domestic rituals
high technology → archival analog systems
minimal brand → dense vernacular signage

The purpose is to escape automatic category aesthetics and expose unexpected relationships.


4. Generate search paths

For each promising territory, generate several queries.

Vary the query type.

For example:

<subject> photography
<subject> archive
<subject> diagram
<subject> technical drawing
<subject> manufacturing
<subject> packaging
<subject> symbol
<subject> icon
<subject> typography
<subject> advertising history
<subject> architecture
<subject> material texture
<subject> cross section
<subject> patent
<subject> machinery
<subject> scientific illustration

Combine terms when useful:

metal joint macro photography
vintage camera focus markings
furniture construction diagrams
robot arm industrial safety graphics
coastal navigation symbols
textile weaving grid structures

Search for the project, not merely for websites about the project.


5. Gather references broadly

Collect enough material to see patterns emerge.

Do not optimize prematurely.

During early collection:

  • Allow variety.
  • Preserve surprising references.
  • Avoid evaluating every reference as a possible interface.
  • Include references from several disciplines.
  • Include both literal and abstract material.
  • Record why a reference was retained when that reason is not obvious.

A large reference pool can be useful, but do not treat a fixed number as the goal.

Use research saturation as the stop condition:

Continue until important visual patterns repeat and new references stop producing materially different insights.

For broad projects, one hundred or more references may be reasonable.

For narrow or well-defined projects, fewer high-signal references may be sufficient.

Quality and diversity matter more than satisfying a numeric quota.


6. Separate references from inspiration targets

For every high-value reference, ask:

What principle am I extracting from this?

Possible answers include:

  • Repeated circular fasteners
  • Extreme material close-ups
  • Layered translucent surfaces
  • Vertical measurement markings
  • Dense modular labeling
  • Alternation between macro and micro scale
  • Cropped portrait framing
  • Strong negative space
  • Sequential process imagery
  • Mechanical rotational movement

Avoid conclusions such as:

Use this exact layout.
Use these colors.
Make the website look like this.
Copy this animation.

Extract transferable principles instead of reproducing identifiable work.


7. Cluster recurring patterns

Group the strongest references by recurring characteristics.

Possible cluster dimensions include:

Form

  • Circular
  • Angular
  • Modular
  • Organic
  • Fragmented
  • Layered
  • Continuous
  • Repetitive

Material

  • Rough
  • Polished
  • Reflective
  • Transparent
  • Fibrous
  • Printed
  • Mechanical
  • Handmade

Composition

  • Dense
  • Sparse
  • Symmetrical
  • Asymmetrical
  • Framed
  • Cropped
  • Layered
  • Sequential

Image behavior

  • Macro
  • Documentary
  • Diagrammatic
  • Portrait
  • Process-based
  • Environmental
  • Archival
  • Abstract

Typography and notation

  • Labels
  • Measurements
  • Serial numbers
  • Captions
  • Editorial display type
  • Handwritten marks
  • Technical notation
  • Wayfinding

Movement

  • Rotation
  • Assembly
  • Expansion
  • Focus
  • Scanning
  • Sliding
  • Folding
  • Melting
  • Connecting
  • Fragmenting

Do not force every reference into a cluster.

Outliers may reveal the most original direction.


8. Identify visual atoms

Extract small recurring ingredients that could eventually become part of a visual language.

A visual atom might be:

  • A shape
  • A border treatment
  • A material
  • A framing device
  • A crop behavior
  • A marker
  • A line
  • A measurement
  • A repeated spatial relationship
  • A motion behavior

For every visual atom record:

Atom:
Observed in:
Project connection:
Potential meaning:
Possible applications:
Risk:

Example:

Atom:
Circular fastener

Observed in:
Metal sheets, bolts, industrial joints, fabrication equipment

Project connection:
The company manufactures and assembles metal components

Potential meaning:
Joining, precision, construction

Possible applications:
Section markers, loading motion, anchor points, image framing

Risk:
Could become decorative industrial cosplay if overused

9. Derive visual metaphor candidates

A visual metaphor should translate something meaningful about the project into a visual or behavioral idea.

Use this chain:

project property
→ real-world or conceptual source
→ visual behavior
→ intended meaning

Example:

Metal fabrication
→ screws and mechanical fasteners
→ circular markers that rotate as sections connect
→ precision, assembly, transformation

Another:

Film production
→ camera focusing mechanism
→ imagery moving from blur to sharp focus
→ selection, attention, cinematic observation

A metaphor is weak when it could be attached to an unrelated project without changing its meaning.

Reject metaphor candidates that are:

  • Pure decoration
  • Based only on visual similarity
  • Overly literal
  • Difficult to understand
  • Culturally inappropriate
  • Dependent on copying another brand
  • Likely to interfere with usability
  • So elaborate that they dominate the content
  • More interesting than relevant

Prefer subtle metaphors that can survive repeated use.


10. Identify category clichés

Review competitors and mainstream category references separately.

Ask:

  • Which layouts recur constantly?
  • Which colors dominate?
  • Which typography choices are predictable?
  • Which imagery is interchangeable?
  • Which phrases are repeated?
  • Which visual metaphors are already exhausted?
  • Which interactions have become category defaults?
  • Which current trends appear regardless of the actual brand?

Create an anti-reference list.

Example:

Avoid defaulting to:

- dark gradient + glowing orb for AI
- dashboard screenshots floating in glass cards for SaaS
- generic handshake photography for B2B
- mountain photography for every outdoor-adjacent brand
- luxury serif + beige palette without brand rationale

Do not ban a convention merely because it is common.

A familiar convention may still be correct when it communicates effectively.

The purpose is awareness, not novelty for novelty's sake.


11. Look for project-owned territory

Now identify what appears unusually specific to the project.

Ask:

  • What references could competitors not credibly claim?
  • Which materials belong to this product?
  • Which processes are visually distinctive?
  • Which historical elements are genuinely ownable?
  • Which words or symbols recur naturally?
  • Which shapes appear repeatedly without being invented by the designer?
  • What could become recognizable after repeated use?
  • What does the project's audience already understand?
  • What has emotional or narrative significance?

These signals are generally stronger foundations than unrelated aesthetic trends.


12. Build candidate graphic universes

Synthesize the research into two or three genuinely different candidate universes.

Do not create superficial variations of the same aesthetic.

Each candidate must contain:

Universe title

Use a short descriptive title.

Avoid vague labels such as:

Modern
Bold
Premium
Minimal
Option A

Prefer titles that describe the underlying logic.

Thesis

One sentence explaining the visual world.

Project origin

Explain which project-specific evidence led to the direction.

Source territories

List the major research territories feeding it.

Dominant forms

Describe recurring geometry and spatial behavior.

Material language

Describe important surfaces, textures, physical qualities, or digital equivalents.

Image direction

Describe subject matter, framing, scale, lighting, documentation style, or abstraction.

Typographic implications

Describe what typography may need to express.

Do not prescribe final fonts unless explicitly requested.

Graphic devices

Describe recurring frames, marks, patterns, symbols, labels, or spatial devices.

Motion and interaction implications

Describe only conceptually meaningful behavior.

Examples:

focus
assembly
rotation
reveal
compression
expansion
layering
sequencing
tracking

Do not design detailed interaction choreography yet.

Signature metaphor

Name the strongest metaphor candidate associated with the universe.

Emotional character

Use precise language.

Prefer:

engineered but not sterile
documentary and tactile
quietly authoritative
imperfect and human
ritualistic and sequential

over:

cool
modern
beautiful
premium

Distinctiveness argument

Explain why this direction belongs to this project rather than merely being fashionable.

Risks

Identify where the direction could become:

  • Gimmicky
  • Too literal
  • Too abstract
  • Difficult to scale
  • Content-dependent
  • Inaccessible
  • Expensive to produce
  • Visually repetitive

Do not select the final creative direction unless the user explicitly asks.

The output should prepare material for a later design-direction decision.


13. Test the universes for extensibility

For each candidate, mentally test whether its logic could influence:

  • Hero
  • Editorial sections
  • Product or service presentation
  • Data or statistics
  • Testimonials
  • Navigation
  • Calls to action
  • Image treatment
  • Motion
  • Mobile layouts
  • Campaign material
  • Social assets

Do not design these artifacts.

Ask only whether the visual language has enough depth to support them.

Reject directions that work only as a hero image.


14. Run the specificity test

Ask:

Could this exact graphic universe be presented to a direct competitor with almost no changes?

If yes, it is not specific enough.

Return to the research.

Look deeper into:

  • History
  • Process
  • Materials
  • Product behavior
  • Audience
  • Founder
  • Location
  • Language
  • Culture
  • Internal terminology
  • Physical artifacts

Distinctiveness should emerge from understanding, not arbitrary weirdness.


15. Run the coherence test

Check whether the proposed universe can explain its own decisions.

For major recommendations, verify:

Why this form?
Why this image treatment?
Why this material?
Why this metaphor?
Why this movement?
Why does it matter to this project?

If the answer is primarily:

because it looks interesting

the recommendation is not sufficiently grounded.


16. Run the restraint test

A graphic universe is a source of possibilities, not an instruction to use everything simultaneously.

Separate findings into:

Core signals

The few ideas most responsible for recognition.

Supporting signals

Elements that reinforce the universe when appropriate.

Experimental signals

Ideas worth testing but not yet foundational.

Prefer a small number of strong recurring principles over a collection of unrelated tricks.


Output format

Return:

Graphic Universe Summary

Project:
Audience:
Core communication goal:
Research question:
Current category problem:

Project Vocabulary

Signal Type Why it matters

Research Territories

Territory Why investigate it Example searches Signal strength

Use signal strength:

high
medium
exploratory

Reference Clusters

Cluster Recurring observations Project relevance Possible implication

Category Clichés

Convention Why it is common Keep, reinterpret, or avoid Reason

Visual Atoms

Atom Evidence Meaning Possible applications Risk

Visual Metaphor Candidates

Metaphor Project origin Visual behavior Intended meaning Risk

Candidate Graphic Universe A

Title:
Thesis:
Project origin:
Source territories:
Dominant forms:
Material language:
Image direction:
Typographic implications:
Graphic devices:
Motion implications:
Signature metaphor:
Emotional character:
Distinctiveness:
Risks:

Candidate Graphic Universe B

Use the same structure.

Candidate Graphic Universe C

Use the same structure when a third meaningful direction exists.

Do not invent a weak third direction merely to fill the template.

Project-Owned Opportunities

Identify the strongest ideas competitors would struggle to claim authentically.

Research Gaps

List unanswered questions or missing source material that could materially change the visual direction.

Handoff

Finish with:

Strongest recurring signals:
Most promising metaphors:
Directions ready for concept exploration:
Ideas that should not proceed:
Open questions:
Recommended next skill:

Evidence discipline

Separate three kinds of statements:

Observed

Something directly visible or documented in project material or external research.

Observed: circular fasteners recur across the company's fabrication imagery.

Inferred

A reasonable interpretation of the observed material.

Inferred: joining and assembly may be useful conceptual themes.

Proposed

A possible design application.

Proposed: circular anchors could become recurring section markers.

Do not present inference as project fact.

When external research is used, preserve useful source references or URLs where the environment supports them.


Research saturation rule

Do not stop after the first attractive cluster.

Continue research while additional sources are still producing meaningful new:

  • Forms
  • Materials
  • Behaviors
  • Historical connections
  • Metaphors
  • Contrasts
  • Structural ideas

Stop when new material mainly repeats already-understood patterns.

A large board is useful only when it increases understanding.


Anti-copy rule

References are evidence and inspiration, not templates.

Do not:

  • Reproduce another designer's composition.
  • Recreate a distinctive branded graphic system.
  • Copy proprietary illustrations.
  • Clone recognizable interactions.
  • Transfer a metaphor that belongs specifically to another brand.

Instead extract principles such as:

scale contrast
modular repetition
mechanical rotation
layered transparency
archival labeling
asymmetric framing

Then reconnect those principles to the current project's own material.


Accessibility and usability guardrail

This is an upstream creative-direction skill, but proposed visual ideas must remain compatible with usable interfaces.

Flag metaphor or art-direction ideas that are likely to require:

  • Distorted reading order
  • Low contrast
  • Text embedded only inside imagery
  • Motion required for comprehension
  • Pointer-only interaction
  • Excessive cognitive load
  • Unreadable typography
  • Disorienting animation
  • Important information hidden behind decorative behavior

Do not prematurely flatten expressive ideas because they may require careful implementation.

Instead distinguish:

Concept is viable with constraints

from:

Concept inherently conflicts with comprehension or access

Quality bar

The task is complete only when:

  • The research begins from the specific project rather than a preselected aesthetic.
  • Current web-design references do not dominate the reference pool.
  • Multiple research territories outside interface design have been explored.
  • Direct, abstract, historical, and adjacent references have been considered where relevant.
  • Important reference clusters are described as principles rather than copy targets.
  • Category conventions and clichés have been identified.
  • The output contains evidence of repeated patterns rather than arbitrary stylistic choices.
  • Visual atoms are traceable to project material.
  • Metaphor candidates have a clear project origin and intended meaning.
  • Weak decorative metaphors have been rejected.
  • Candidate graphic universes differ conceptually rather than only by color or typography.
  • At least one candidate explores territory that would not normally appear in a standard competitor moodboard.
  • Every candidate can extend beyond a single hero treatment.
  • Recommendations distinguish observation, inference, and proposal.
  • Copyright-sensitive references are treated as principles, not templates.
  • Accessibility and comprehension risks are identified early.
  • Research stops because saturation has been reached, not because an arbitrary reference count was completed.
  • The result gives the next design workflow meaningful material to work with.

Edge cases

Thin or incomplete brief

If little project information exists:

  1. Create a provisional vocabulary.
  2. Mark assumptions clearly.
  3. Explore only low-risk territories.
  4. Produce research questions alongside findings.
  5. Do not claim the resulting universe is project-specific until the assumptions are validated.

Existing strong brand identity

Use existing brand assets as research anchors.

Do not search for a replacement aesthetic unless rebranding is explicitly in scope.

Investigate how existing identity elements can connect to:

  • Product
  • Content
  • Environment
  • Movement
  • Layout
  • Photography
  • Digital behavior

Visually "boring" B2B project

Do not conclude that there is no visual material.

Investigate:

  • Tools
  • Facilities
  • Workflows
  • Documents
  • Machinery
  • Materials
  • Technical language
  • Diagrams
  • Logistics
  • Processes
  • Transformation

Operational detail often provides more distinctive material than competitor websites.

Purely digital product

If there is no physical artifact, investigate:

  • User behavior
  • Information flow
  • Data relationships
  • Transformation
  • Inputs and outputs
  • Collaboration patterns
  • Language
  • States
  • Timing
  • Feedback loops
  • System architecture at a conceptual level

Do not default automatically to abstract gradients and glowing geometry.

User provides only same-style references

Treat them as preference evidence, not the entire research universe.

Identify the common qualities the user appears to prefer, then deliberately explore other territories that express those qualities through different visual sources.

No network access

Work from supplied project material and references.

Still produce:

  • Project vocabulary
  • Research territory map
  • Search query matrix
  • Preliminary patterns
  • Research gaps

Clearly identify external research that remains to be performed.

Culturally specific references

Do not use symbols, rituals, historical imagery, indigenous patterns, religious material, or culturally specific visual languages solely for aesthetic effect.

Verify meaning and context before recommending them.

Too many unrelated findings

Return to the project vocabulary.

Prioritize clusters with the strongest combination of:

project relevance
+ recurrence
+ distinctiveness
+ extensibility

Discard interesting but irrelevant material.


Related skills

Use before this skill when available:

  • design-problem-framer — clarifies the actual design problem, audience, business transformation, and communication objectives.

Use after this skill when available:

  • story-led-website-concept (proposed) — turns strategic and visual research into narrative-led website concepts.
  • design-direction-decision — compares mature creative directions and recommends which should proceed.
  • brand-derived-visual-system (proposed) — converts a chosen direction into a reusable visual grammar.
  • expressive-grid-planner (proposed) — derives layout structures appropriate to the chosen content and art direction.
  • art-direction-coherence-review (proposed) — checks whether content, typography, imagery, layout, motion, and interface language support the same intended meaning.
  • layout-principles-review — reviews hierarchy, proportion, rhythm, alignment, balance, and composition.

Do not duplicate these downstream jobs inside this skill.

The graphic-universe-builder should finish with strong research material and visual-direction seeds, not a finished website.

The main distinction is important: this skill stops before concept selection and visual-system design. That keeps it narrow enough to route independently, which matches the library's “one clear reusable job” rule.