Image Generation

Image Generation

Generate images from text descriptions using AI image models. This skill guides you through constructing effective image generation requests, selecting appropriate providers, and presenting results to the user.

When to Use

Use the image generation tools when the user:

• Explicitly asks for an image, illustration, diagram, or visual asset
• Needs a hero image, banner, icon, or other visual for a project
• Describes a scene, concept, or design they want visualised
• Asks you to "show me", "draw", "create a picture of", or similar

Do NOT generate images unless the user expresses a clear intent for visual output. Do not assume a text description is an image request unless context makes it obvious.

Available Tools

list_image_models

Discover which image providers are configured and what models/capabilities are available. Call this tool first when:

• You are unsure which providers the user has set up
• You need to check available models before suggesting one
• The user asks about supported formats, sizes, or quality levels

This tool works without API keys.

create_visual_asset

Generate an image from a text prompt. Returns a local file path, file:// URL, and metadata.

Required parameter: prompt (the image description)

Optional parameters:

Parameter	Default	When to change
provider	auto	User requests a specific provider
model	Provider default	User requests a specific model
aspect_ratio	1:1	Match the intended use case (see below)
quality	auto	User needs faster generation (low) or higher detail (high)
output_format	png	Use jpeg for photos, webp for web assets
filename_hint	Generated	Provide a descriptive name for the output file

Constructing Effective Prompts

Build prompts that cover these dimensions:

1. Subject: What is the main content? Be specific.
2. Style: Photorealistic, illustration, watercolour, flat design, 3D render, etc.
3. Composition: Close-up, wide shot, bird's-eye view, centred, rule-of-thirds, etc.
4. Mood/lighting: Warm, cool, dramatic, soft, natural daylight, studio lighting, etc.
5. Context/use: Where will the image be used? This informs aspect ratio and style.

Example prompt construction

User says: "I need a hero image for my AI notebook app"

Constructed prompt:

A clean, modern hero image for a privacy-focused AI notebook application. Minimalist composition with a softly glowing notebook icon in the centre, surrounded by abstract neural network patterns. Cool blue and white colour palette with subtle gradients. Professional, trustworthy aesthetic suitable for a landing page.

Parameters: aspect_ratio: "16:9", quality: "high", filename_hint: "ai_notebook_hero"

Choosing Aspect Ratios

Use case	Recommended ratio
Social media post, avatar, icon	1:1
Website hero, banner, presentation slide	16:9
Mobile wallpaper, story, vertical banner	9:16
Blog thumbnail, product card	4:3
Portrait, book cover	3:4 or 2:3
Ultra-wide banner, cinematic	21:9

Provider Selection Strategy

• auto (default): Let the system choose the best available provider. This is correct for most requests.
• openai: Prefer when high quality and detailed text rendering matter. Model gpt-image-2 supports transparent backgrounds.
• gemini: Prefer for fast generation or when the user specifically requests Gemini models.

Before requesting a specific provider, call list_image_models to verify it is configured.

Presenting Results

When image generation succeeds:

1. Show the file path so the user can locate the file
2. Mention the provider and model used
3. If the aspect ratio or quality differs from what the user might expect, explain why
4. Offer to regenerate with different parameters if the user is not satisfied

Example:

Image generated successfully using OpenAI gpt-image-2. File: /path/to/generated-images/ai_notebook_hero_20260602_143000_a1b2c3.png

The image uses a 16:9 aspect ratio at high quality. Let me know if you would like me to adjust the style, composition, or try a different prompt.

Handling Errors

Error code	What to tell the user
NO_PROVIDER_CONFIGURED	No image provider is set up. Ask them to configure an API key in plugin settings or set the environment variable.
PROVIDER_NOT_CONFIGURED	The requested provider is not configured. Suggest trying auto or configuring the missing key.
MODEL_NOT_SUPPORTED	The requested model is not available. Call list_image_models and suggest an alternative.
INVALID_INPUT	A parameter is invalid. Check the error details and fix the input.
GENERATION_FAILED	The provider API call failed. Suggest retrying, simplifying the prompt, or trying a different provider.
GENERATION_TIMEOUT	The request timed out. Suggest reducing quality or trying a different provider.
STORAGE_FAILED	Could not save the image file. Ask the user to check the output directory permissions.

Do not guess what went wrong. Read the error message and suggestion, then relay actionable advice to the user.