image-cog

AI image generation powered by CellCog. Create images, edit photos, consistent characters, product photography, reference-based images, sets of images, style...

Security Scan

The skill is an instruction-only wrapper that delegates image generation to the separate 'cellcog' skill and is internally consistent, but you should inspect the referenced cellcog skill (credentials, routing, and data handling) before installing.

Purpose & Capability

The skill's name/description (AI image generation) matches the instructions: it tells the agent how to call CellCog SDK to create/edit images and describes supported image types and models. It does not request unrelated capabilities.

Instruction Scope

SKILL.md does not instruct the agent to read local files, system paths, or unrelated environment variables. However, it explicitly requires the separate 'cellcog' skill for SDK setup and API calls—so runtime behavior (network endpoints, credentials, routing) is governed by that dependent skill. You should review cellcog's instructions to confirm no surprising scope creep there (e.g., reading shell history or arbitrary files).

Install Mechanism

This is instruction-only with no install spec or code files, so nothing gets downloaded or written by this skill itself. The SKILL.md suggests 'clawhub install cellcog'—the install risk depends on the cellcog skill, not this one.

Credentials

This skill declares no required environment variables or credentials. That is consistent only if the dependent 'cellcog' skill handles all API keys and routing. The SKILL.md names external model providers (Google, OpenAI, Recraft); verify whether you must supply provider API keys or whether cellcog will proxy requests using its own credentials (and what that implies for cost, rate limits, and data privacy).

Persistence & Privilege

The skill does not request always:true and leaves model invocation enabled (normal). It does not indicate any system-wide configuration changes or persistent elevated privileges.

Assessment

This skill itself is a thin instruction wrapper that delegates all work to the 'cellcog' skill. Before installing, review the cellcog skill (its SKILL.md or source) to confirm: 1) which environment variables or API keys it requires and where requests are sent; 2) whether it proxies requests using its own credentials (and whether you trust that provider/operator); 3) any install scripts or downloads performed by cellcog; 4) how it handles uploaded images and whether the data is retained/shared. If you cannot inspect cellcog or its install sources, treat installation as higher risk. Also check costs and rate limits for the named providers (Google/OpenAI/Recraft) so unexpected billing doesn't occur.

Image Cog - AI Image Generation Powered by CellCog

Create professional images with AI - from single images to consistent character sets to product photography.

---

Prerequisites

This skill requires the cellcog skill for SDK setup and API calls.

clawhub install cellcog

Read the cellcog skill first for SDK setup. This skill shows you what's possible.

Quick pattern (v1.0+):

# Fire-and-forget - returns immediately
result = client.create_chat(
    prompt="[your image request]",
    notify_session_key="agent:main:main",
    task_label="image-task",
    chat_mode="agent"  # Use "agent" for simple images, "agent team" for complex
)
# Daemon notifies you when complete - do NOT poll

---

What Models Do We Use

Model	Provider	Primary Use
Nano Banana 2 (Gemini 3.1 Flash Image)	Google	Default image generation — photorealistic scenes, complex compositions, text rendering, multi-turn character consistency
GPT Image 1.5	OpenAI	Transparent background images — logos, stickers, product cutouts, overlay graphics
Recraft	Recraft AI	Scalable vector illustrations (SVG) and icon generation

Nano Banana 2 is the default model for all image generation. CellCog's agents intelligently route to other models when the task calls for it — for example, transparent PNGs are automatically handled by GPT Image 1.5, and vector/icon requests go to Recraft. If you'd prefer a specific model, just mention it in your prompt (e.g., "use ChatGPT/OpenAI image generation").

What Images You Can Create

Single Image Creation

Generate any image from a text description:

• Scenes: "A cozy coffee shop interior with morning light streaming through windows"
• Portraits: "Professional headshot of a confident woman in business attire"
• Products: "Minimalist product shot of a white sneaker on a marble surface"
• Abstract: "Geometric abstract art in navy and gold"
• Nature: "Misty mountain landscape at sunrise with a lone hiker"

Image Editing

Transform existing images:

• Style Transfer: "Transform this photo into a watercolor painting"
• Background Removal: "Remove the background and place on a clean white backdrop"
• Enhancement: "Enhance the colors and add dramatic lighting"
• Modification: "Change the person's outfit to a red dress"

Consistent Characters

Create multiple images of the same character in different scenarios:

• Character Series: "Create a tech entrepreneur character, then show them: 1) At their desk coding, 2) Presenting to investors, 3) Celebrating a product launch"
• Mascot Variations: "Design a friendly robot mascot, then create versions for: welcome page, error page, success message, loading screen"
• Story Sequences: "Create a main character, then illustrate them in 5 scenes of a journey"

This is powerful for:

• Comic strips and storyboards
• Marketing campaigns with consistent characters
• Video frame generation
• Brand mascots across contexts

Product Photography Style

Professional product visuals:

• Hero Shots: "Product hero shot of a smartwatch on a gradient background"
• Lifestyle Shots: "Smartphone being used by a person in a modern living room"
• Flat Lays: "Flat lay of skincare products with botanical elements"
• 360 Views: "Multiple angles of a leather handbag - front, side, back, detail"

Sets of Related Images

Multiple cohesive images for campaigns or collections:

• Social Media Sets: "5 Instagram post images for a fitness brand - consistent style, varied content"
• Website Heroes: "3 hero images for a SaaS landing page - professional, modern, tech-focused"
• Ad Variations: "4 versions of a product ad with different backgrounds and moods"
• Blog Illustrations: "Set of 6 illustrations for a blog post about productivity tips"

Reference-Based Generation

Use existing images as references for style, character, or composition:

• Style Matching: "Create a new image in the same artistic style as this reference"
• Character Consistency: "Using this person as reference, create a new scene with them hiking"
• Brand Alignment: "Create product images matching this brand's visual style"
• Composition Reference: "Create a similar composition but with different subjects"

---

Image Specifications

Aspect	Options
Aspect Ratios	1:1 (square), 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9
Sizes	1K (~1024px), 2K (~2048px), 4K (~4096px)
Styles	Photorealistic, illustration, watercolor, oil painting, anime, digital art, vector
Formats	PNG (default)

Size recommendations:

• 1K: Quick iterations, thumbnails, social media posts, drafts
• 2K: Standard web content, presentations, marketing materials
• 4K: Hero images, print materials, final deliverables where detail matters

---

When to Use Agent Team Mode

For image generation, chat_mode="agent team" is recommended for:

• Complex scenes requiring multiple elements
• Consistent character series
• Reference-based generation requiring analysis
• Sets of related images

For simple single images, chat_mode="agent" can work faster.

---

Example Image Prompts

Professional headshot:

"Create a professional headshot of a friendly Asian woman in her 30s, wearing a navy blazer, soft studio lighting, neutral gray background, confident but approachable expression. 1:1 square, 2K quality, photorealistic."

Product photography:

"Product shot of a premium wireless earbuds case, matte black finish, on a reflective dark surface with subtle blue accent lighting. Minimalist, high-end tech aesthetic. 4:3 landscape, 4K for hero image."

Consistent character set:

"Create a character: young Black male software developer, casual style with glasses, friendly demeanor. Then create 4 images:

1. Working at a standing desk with multiple monitors
2. In a video call meeting, explaining something
3. At a coffee shop with laptop, thinking
4. Celebrating with team, high-fiving Keep the character exactly consistent across all images."

Social media set:

"Create 5 Instagram posts for a plant-based meal delivery service:

1. Colorful Buddha bowl from above
2. Happy person unpacking delivery
3. Meal prep containers arranged neatly
4. Close-up of fresh ingredients
5. Before/after showing ingredients to finished dish Style: bright, fresh, appetizing, consistent warm color grading. 1:1 square format."

Style transfer:

"Transform this uploaded photo of a city street into a Studio Ghibli anime style illustration. Keep the composition and elements but apply the characteristic Ghibli warmth, soft clouds, and whimsical details."

---

Tips for Better Images

1. Be descriptive: "Woman in office" is vague. "Confident woman in her 40s, silver blazer, modern glass-walled office, warm afternoon light" is better.

2. Specify style: "Photorealistic", "digital illustration", "watercolor", "minimalist vector".

3. Describe lighting: "Soft natural light", "dramatic side lighting", "golden hour glow", "studio lighting".

4. Include mood: "Professional and confident", "warm and inviting", "energetic and vibrant".

5. Mention composition: "Rule of thirds", "centered symmetry", "close-up", "wide establishing shot".

6. For consistency: When creating character series, describe the character in detail first, then reference "the same character" in subsequent prompts.