visual-creation

Visual Creation

Generate on-brand visuals using a defined style system, not vibes.

How This Skill Works

1. Load style DNA — pull visual rules from CREATOR.md or creator-memory/visual.md
2. Engineer prompt — translate concept into a detailed image generation prompt
3. Apply quality gates — check against brand standards before accepting

If no visual style is documented: prompt the user, or run /creator-sprint.

Visual Style DNA

Before generating, confirm these parameters:

Parameter	What to capture	Example
Color palette	Primary + 2-4 accents	Deep navy, cyan, gold, white
Mood/aesthetic	Overall feeling	Dark premium, editorial, clean
Subject preference	What appears	Technology, abstract, humans
Composition	How elements arrange	Rule of thirds, centered, dynamic
Style reference	Visual category	Glassmorphic, editorial photo, 3D
Anti-patterns	What to avoid	Flat clipart, chrome robots

Prompt Engineering

A good image generation prompt has 6 components:

[Style/Medium] [Subject] [Setting/Context] [Mood] [Color Direction] [Technical Quality]

Example — dark premium tech:

Dark premium 3D render, glowing holographic interface floating above sleek desk surface,
dramatic cyan and gold lighting against deep navy background, depth of field blur on
background elements, cinematic composition, highly detailed, 8K quality, no text, no humans

Example — light editorial:

Clean editorial photograph, open laptop with code on screen, minimalist desk setup,
soft morning light from left, white and warm tones, shallow depth of field, lifestyle
photography style, high resolution

Style modifiers by type

Style	Key modifiers
Dark premium	dark background, dramatic lighting, deep shadows, cinematic
Light editorial	white/light tones, natural light, minimal, lifestyle photography
3D render	3D render, subsurface scattering, ray tracing, realistic materials
Glassmorphic	frosted glass, transparency, blur, depth layers, neon accents
Cinematic	film grain, dramatic color grading, shallow depth of field

7-Gate Quality Filter

All gates must pass before accepting an image:

Gate	Check	Fail condition
1. Brand alignment	Colors match palette	Wrong colors, off-brand feel
2. Color balance	3+ distinct tones	Single-color saturation
3. Composition depth	Foreground + mid + background	Flat, no depth
4. Subject quality	Main subject is clear	Ambiguous, cluttered
5. Text legibility	If text exists, it's readable	Blurry, wrong font style
6. Anti-pattern check	None of the vetoed elements	Contains banned elements
7. Scroll-stop	Would this stop scrolling?	Generic, forgettable

Image Types

Type	Dimensions	Key requirements
Blog hero	1200×628	No text, scroll-stop, brand colors
Social (LinkedIn)	1200×627	Works with text overlay, high contrast
Social (X)	1600×900	Bold composition, clear focal point
Newsletter header	1200×400	Wide, simple
Thumbnail	1280×720	High contrast, small-size readable
Product hero	1:1 or 4:5	Clean, premium, minimal

Output Format

Deliver:

1. Prompt — ready to paste into generation tool
2. Negative prompt — elements to exclude
3. Rationale — key choices explained
4. Alternatives — 1-2 prompt variations

After generation: apply gates, report pass/fail, suggest adjustments on fail.

Uses ~~image generation. See references/prompt-library.md for tested prompts by category.