Skip to content
OpenAgentSkill
Browsing is open · Catalog captured 2026-11-21 · Automatic updates and account services are paused.
Marketing & sales

creative-direction

Image prompt templates, model selection guidance, and anti-generic patterns for generating visual assets. Use when the user needs AI-generated images for landing pages, marketing, or products. Covers hero images, feature illustrations, OG cards, icons, and backgrounds.

dylanfeltus · Static checks on file

Installation

Read the source and dependencies before running a command. Copying does not install or execute anything.

npx skills add dylanfeltus/skills --skill creative-direction

About this skill


name: creative-direction description: Image prompt templates, model selection guidance, and anti-generic patterns for generating visual assets. Use when the user needs AI-generated images for landing pages, marketing, or products. Covers hero images, feature illustrations, OG cards, icons, and backgrounds.

Creative Direction

Image prompt templates, model selection guidance, and anti-generic patterns for generating visual assets. Covers hero images, feature illustrations, testimonial photos, OG images, and more.

When to Use

  • User needs images for a landing page, app, or marketing site
  • User asks for "hero image", "feature illustration", or "OG image"
  • User wants AI-generated visuals that don't look like stock photos
  • User is choosing between image generation models
  • User's current AI images look generic and need direction

Core Philosophy

  • Generic is the enemy. "A person using a laptop" produces forgettable images. Specificity creates memorability.
  • Every image has a job. Hero = emotion + aspiration. Feature = clarity. Testimonial = trust. Know the job before prompting.
  • Model matters. Different models excel at different things. Pick the right tool.
  • Consistency beats quality. A cohesive set of 7/10 images beats one 10/10 with six mismatches.

Model Selection Guide

When to Use Each Model

ModelBest ForWeaknessesCost
GPT-4o / DALL-E 3Text in images, diagrams, infographics, precise compositionsCan feel "illustrated", less photorealisticAPI credits
Gemini ImagenPhotorealism, natural scenes, product shotsLess control over composition, text rendering variesAPI credits
MidjourneyArtistic quality, mood, cinematic shots, brand imageryNo API (Discord-only), inconsistent with specific detailsSubscription
Flux (via Replicate)Photorealism, faces, flexible stylesRequires Replicate accountPer-image
Unsplash / PexelsReal photography, when AI looks too AILimited to what exists, generic posesFree

Decision Tree

Need text in the image? → DALL-E 3 / GPT-4o
Need photorealistic people? → Flux or Gemini
Need artistic/cinematic mood? → Midjourney
Need a specific real-world scene? → Unsplash/Pexels
Need consistency across many images? → Pick ONE model, same style prompt prefix
Need diagrams/UI mockups? → DALL-E 3 / GPT-4o

Prompt Templates by Asset Type

Hero Image

Job: Create an emotional first impression. Communicate the product's vibe in 2 seconds.

Template:

[Art style], [subject doing something specific and aspirational],
[environment with mood-setting details], [lighting description],
[color palette constraint], [composition note]

Example — SaaS Product:

Cinematic photograph, a product designer reviewing a clean dashboard
on a large monitor in a sunlit corner office, morning golden hour
light casting long shadows, muted blue and warm cream color palette,
shot from over the shoulder with shallow depth of field, 35mm lens feel

Anti-generic patterns:

  • ❌ "A person using a computer" → ✅ "A designer reviewing analytics on a ultrawide monitor, sticky notes scattered on the desk"
  • ❌ "Happy team working" → ✅ "Three engineers around a whiteboard, one mid-laugh, marker in hand, late afternoon light"
  • ❌ "Technology abstract" → ✅ "Closeup of hands arranging physical cards on a table, each card showing a tiny wireframe sketch"

Feature/Product Illustration

Job: Explain what a feature does. Clarity > beauty.

Template:

[Clean/minimal style], [the feature's core concept as a visual metaphor],
[simple background], [brand colors], [no text unless needed]

Example — Analytics Feature:

Minimal 3D illustration, a translucent glass cube containing floating
data points that form a rising trend line, soft gradient background
from white to light blue, subtle shadows, isometric perspective

Tips:

  • Use visual metaphors, not literal screenshots
  • Keep backgrounds simple — the illustration should work on any page section
  • Maintain consistent style across all feature illustrations (same rendering style, same perspective, same color treatment)

Testimonial / Social Proof

Job: Build trust. Make real people feel real.

Approach: Use real photos when possible (with permission). If generating:

Professional headshot, [specific person description with age/style details],
[neutral or office background], [natural lighting], [warm but professional mood],
shot at eye level, slight smile, [avoid uncanny valley — add imperfections]

Tips:

  • Diversity matters — vary age, ethnicity, style
  • Avoid the "corporate headshot" look — slightly candid feels more trustworthy
  • ⚠️ AI-generated faces for testimonials is ethically questionable. Prefer real photos. If generating, be transparent about it.

Open Graph (OG) / Social Card

Job: Get clicks in a feed. Must work at small sizes.

Template:

[Bold, high contrast], [simple central element],
[large readable text area (left or center)],
[brand color background or gradient], [16:9 aspect ratio],
minimal detail — this will be viewed at 600×315px

Key constraints:

  • Must be readable at thumbnail size
  • Text should be generated separately and composited (AI text rendering is unreliable)
  • High contrast between background and text area
  • Simple shapes > complex scenes

Icon / Logo Concept

Job: Convey brand identity in a tiny space.

Minimal vector icon, [object/symbol], [single or two-color],
clean lines, works at 32px, [style: geometric/rounded/sharp],
white background, no gradients, no shadows

Tips:

  • Generate concepts, then recreate in Figma/SVG for production
  • AI-generated logos are starting points, not finals
  • Test at small sizes — if it doesn't read at 32px, simplify

Background / Texture

Job: Add depth without competing with content

Read the complete source ↗