Content Image

Content Image Skill

Important: This Skill Generates Prompts, Not Images

Claude cannot generate images. This skill produces ready-to-paste prompts optimized for AI image generation models. The output is text that the user copies into DALL-E, Midjourney, Stable Diffusion, or another image generation tool.

Every output includes:

1. The prompt text (optimized for the target model)
2. Recommended model and settings
3. Negative prompt (where the model supports it)
4. Variations to try

When This Skill Activates

• User asks for an image, illustration, or visual asset
• User wants to create image prompts or optimize existing prompts
• User mentions DALL-E, Midjourney, Stable Diffusion, Flux, or other image generation tools
• User says "generate an image of..." or "I need a visual for..."
• User wants a hero image, social media graphic, blog illustration, or product shot

Prompt Structure

Every effective image prompt has four components:

1. Subject

What is in the image. Be specific and concrete.

• Bad: "a person working"
• Good: "a software developer sitting at a standing desk with dual monitors, writing code, coffee mug nearby"

The subject should be described in enough detail that two different artists would produce recognizably similar images.

2. Style

The visual treatment. Reference a specific style from the style library or describe it explicitly.

• Bad: "nice looking"
• Good: "editorial photography, shallow depth of field, natural window light, 35mm lens"

See style-library/ for pre-built style descriptions:

• style-library/editorial.md -- Clean, professional photography
• style-library/flat-illustration.md -- Bold, geometric, vector-like
• style-library/3d-render.md -- Soft lighting, smooth surfaces, product-shot feel
• style-library/film-still.md -- Cinematic, dramatic, atmospheric

3. Mood / Atmosphere

The emotional tone of the image.

• Calm, focused, productive
• Energetic, vibrant, optimistic
• Dramatic, tense, high-stakes
• Warm, inviting, approachable
• Minimal, quiet, contemplative

4. Technical Specs

Resolution, aspect ratio, and model-specific parameters.

• Aspect ratio: 16:9 (hero banners), 1:1 (social), 4:5 (Instagram portrait), 3:2 (blog photos)
• Quality: high detail, photorealistic vs. stylized
• Lighting: natural, studio, dramatic, flat, backlighting
• Camera: wide angle, close-up, overhead, eye-level

Generation Flow

Step 1: Read Brand Context

If brand/VISUAL.md exists, read it for:

• Color palette (suggest prompts that align with brand colors)
• Visual style preference (photography vs. illustration, minimal vs. bold)
• Mood and tone (professional, playful, technical)

If the brand aesthetic is "clean, minimal, professional," do not suggest prompts for grungy textures or maximalist compositions.

Step 2: Understand the Need

Ask:

• What is the image for? Blog header, social post, presentation slide, product page, email banner.
• What should it show? The subject matter.
• What feeling should it evoke? The mood.
• Any existing visual references? Links to images they like.
• Target model? DALL-E, Midjourney, Stable Diffusion, or no preference.

If the user provides a vague request ("I need an image for my blog post about AI"), ask for specifics before generating prompts.

Step 3: Select Style

Based on the brand and use case, recommend a style:

Step 4: Build Prompt

Assemble the prompt using the four components (subject + style + mood + technical specs). Structure varies by model -- see model-specifics.md.

Step 5: Add Negative Prompt

For models that support negative prompts (Stable Diffusion, some Midjourney modes), specify what to avoid:

Common negative prompt elements:

• "blurry, low quality, distorted, deformed"
• "text, watermark, signature, logo" (unless text is desired)
• "stock photo, generic, clip art" (for editorial/premium look)
• "oversaturated, neon, garish" (for professional/muted aesthetics)

Step 6: Provide Variations

Always offer 2-3 prompt variations:

1. Close-up variant: Tighter framing, more detail on the subject.
2. Wide variant: Environmental context, more setting.
3. Alternative style: A different visual treatment for comparison.

Brand Alignment

When generating prompts, align with the brand aesthetic from brand/VISUAL.md:

Color Alignment

If the brand uses warm earth tones, suggest prompts with:

• "warm lighting, golden hour, earth tones, amber and terracotta palette"

If the brand uses cool blues and grays, suggest:

• "cool lighting, blue-gray tones, steel and slate palette, overcast sky"

Do not force brand colors into every image, but guide the aesthetic direction.

Style Alignment

Consistency Across Assets

When generating multiple prompts for the same project (e.g., a series of blog post images), maintain consistency:

• Same style and mood across all prompts
• Same aspect ratio and quality settings
• Same color temperature and lighting direction
• Reference the same style library entry for all prompts in the series

Model-Specific Optimization

See model-specifics.md for detailed guidance. Summary:

DALL-E 3

• Uses natural language descriptions
• Longer, more descriptive prompts work well
• Specify style, mood, and technical details in full sentences
• No negative prompt support (describe what you want, not what you do not want)

Midjourney

• Responds well to stylistic keywords and aesthetic references
• Use --ar for aspect ratio, --s for stylization, --q for quality
• Shorter, more keyword-driven prompts can be effective
• Version-specific parameters (v6, v6.1)

Stable Diffusion

• Supports weighted tokens: (important detail:1.3) for emphasis
• Negative prompts are critical for quality
• Sampler and step count affect output
• LoRA and checkpoint selection matter

Output Format

For each prompt request, deliver:

## Prompt

[The full prompt text, ready to paste]

## Model & Settings

- **Recommended model**: [DALL-E 3 / Midjourney v6.1 / SDXL]
- **Aspect ratio**: [16:9 / 1:1 / 4:5 / etc.]
- **Quality/Steps**: [model-specific settings]

## Negative Prompt (if applicable)

[Negative prompt text for SD/MJ]

## Variations

### Variation 1: [Description]
[Alternative prompt]

### Variation 2: [Description]
[Alternative prompt]

## Notes

[Any tips for getting the best result: regenerate if X, adjust Y for Z]

Common Prompt Patterns

The "Person Doing Thing" Pattern

[demographic detail] [person description] [action] in [setting], [style], [lighting], [mood]

Example: "A mid-30s Southeast Asian woman reviewing data on a tablet in a modern open office, editorial photography, natural window light from the left, focused and productive atmosphere"

The "Product Shot" Pattern

[product] on [surface/background], [camera angle], [lighting setup], [style]

Example: "A matte black wireless keyboard on a minimal white desk, 45-degree angle, soft studio lighting with subtle shadow, 3D product render, clean and modern"

The "Abstract Concept" Pattern

[visual metaphor for concept], [style], [color palette], [mood]

Example: "Interconnected glowing nodes forming a neural network pattern, dark background, blue and purple gradient, isometric 3D render, futuristic and clean"

The "Environment" Pattern

[place description], [time of day], [weather/atmosphere], [style], [camera position]

Example: "A coworking space with floor-to-ceiling windows at golden hour, plants on desks, warm natural light, editorial photography, wide angle from the entrance"

What This Skill Does Not Do

• Generate actual images (Claude is a text model)
• Create SVG graphics (use content-infographic for that)
• Design logos or brand marks (use a designer or specialized tool)
• Edit existing photos (use Photoshop, Lightroom, or similar)

If the user needs an SVG or infographic, redirect to the content-infographic skill. If they need a presentation visual, redirect to content-presentation.