Guide · Agent skill
How to use the SEO Image Generator agent skill (seo-image-gen)
Updated · written with AI from the project's own documentation and checked against our catalogue data
seo-image-gen is an agent skill from the claude-seo project that lets an AI agent generate images for search and social use, such as social preview (OG) images, blog hero images, product shots, infographics and Pinterest pins. It is built on Gemini image generation through the banana extension and nanobanana MCP server. This guide explains what the skill does, how to install it, how to ask for images in plain language, and what to check after each image is made.
SEO Image GeneratorAgriciDaniel/claude-seo · MIT Download the skill Quick start
- 1Check the prerequisites. Confirm you have a Google Gemini API key from Google AI Studio and that the banana extension is available in your setup.
- 2Install the banana extension. Run ./extensions/banana/install.sh from the repository root so the nanobanana MCP server can be connected.
- 3Download and unzip the skill. Download the seo-image-gen ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 4Make a request. Ask in plain language, for example for an OG image for a specific article, naming the subject and purpose.
- 5Review the output. Check that the image matches your request and is appropriate for publication, and note any cost estimate shown.
- 6Finish the SEO steps. Write alt text, rename the file to the keyword-description-widthxheight.webp pattern and convert it to WebP.
- 7Add markup. Add the og:image meta tags and, if useful, an ImageObject schema block for the image.
What is an agent skill and what does seo-image-gen do?
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent (such as Claude) reads when a task matches the skill's description. The skill tells the agent how to approach that kind of task, which tools to use and which steps to follow.
The seo-image-gen skill is for generating SEO assets. It maps each common SEO need to preset aspect ratios, resolutions and visual styles. For example, an OG/social preview image uses a 16:9 ratio at 1K, while a Pinterest pin uses 2:3 at 2K. After generation, the skill guides you through alt text, file naming, WebP conversion, ImageObject schema and OG meta tags.
- Social preview (OG) images, blog hero images, schema images, social square images, product photos, infographics, favicons and Pinterest pins
- Post-generation checklist: alt text, SEO-friendly file names, WebP conversion, file size targets
- Suggested ImageObject schema and og:image meta tags
Install the banana extension and the skill
The skill requires the banana extension. Its installer is run from the repository root with ./extensions/banana/install.sh. The skill also needs a nanobanana MCP server connected to Gemini image generation, and a Google Gemini API key from Google AI Studio (https://aistudio.google.com/apikey).
To install the skill itself, download the seo-image-gen ZIP from this page and unzip it into one of these locations:
- Personal: ~/.claude/skills/ (available across your projects)
- Project: .claude/skills/ inside a specific project folder
- Claude apps: upload the ZIP under Settings → Capabilities → Skills
./extensions/banana/install.shTrigger the skill with natural requests
You do not need to memorise commands. The skill is meant to activate when you ask for an SEO asset using terms such as "OG image", "social preview", "hero image", "blog image", "product photo", "infographic", "seo image", "image-gen", "favicon", "schema image" or "pinterest pin". Requests for general artwork, banners or thumbnails unrelated to SEO are outside its scope.
The skill's source also lists command forms that take a mode and a description: og, hero, product, infographic, custom and batch. Batch generates N variations, with a default of 3.
- "Make an OG image for my article on remote team onboarding."
- "Create a blog hero image for a post about composting at home."
- "I need a product photo of a ceramic mug on a white background for my shop page."
- "Design an infographic showing our three-step returns process."
- "Generate a Pinterest pin for my sourdough recipe post."
/seo image-gen og <description>What the agent does for each request
For every generation, the agent identifies the use case, applies the matching SEO defaults, sets the aspect ratio, builds a detailed prompt and calls the image tool. The source states that cost is shown before generating, especially for batch runs, and that each generation is logged.
Default settings from the skill's use-case table:
- OG/Social Preview: 16:9, 1K
- Blog Hero: 16:9, 2K
- Schema Image: 4:3, 1K
- Social Square: 1:1, 1K
- Product Photo: 4:3, 2K
- Infographic: 2:3, 4K
- Favicon/Icon: 1:1, 512
- Pinterest Pin: 2:3, 2K
Finish the image for SEO after generation
Generating the image is only half the job. The skill's checklist asks you to write descriptive alt text that names the subject, rename the file in the form keyword-description-widthxheight.webp, and convert to WebP for speed. The source suggests a target of under 200KB for hero images and under 100KB for thumbnails.
For social previews, add og:image, og:image:width, og:image:height and og:image:alt meta tags, and consider an ImageObject schema block that lists the URL, width, height and caption.
- Alt text: describe the subject of the image
- File naming: keyword-description-widthxheight.webp
- WebP conversion with ImageMagick: magick output.png -quality 85 output.webp
- Schema: ImageObject with url, width, height and caption
- Open Graph tags for social preview images
magick output.png -quality 85 output.webpSafety, cost and network notes
The skill runs scripts and uses the network. Image generation calls Gemini through the MCP server, and the source says each generation is billed. Pricing is not stored in the skill; the source directs you to Google's pricing page and to a local file, ~/.banana/pricing.json, for dated values, and says estimates are approximate.
If an image is rejected by safety filtering (IMAGE_SAFETY), the source suggests rephrasing the prompt. If you hit a rate limit (429), wait 60 seconds and retry, and check current free-tier limits before running batches.
- Review the cost estimate before large batch runs
- Do not paste API keys into shared chats or commit them to repositories
- Check the generated image against reality before publishing, since AI output may not match a real product or person
Tips for better results
Name the subject, setting and purpose in your request. The skill builds its prompt from your description, so a specific description produces more useful output. Mention the page or post the image is for so the agent can choose the right ratio and mode.
If you already have an SEO audit, the skill is designed to work with image findings: the source says results from the seo-images analysis can drive image-gen commands.
- Say where the image will appear (article, product page, Pinterest)
- Specify the subject and style you want in plain words
- Ask for alt text and file name suggestions in the same request
What people use it for
- Create a 16:9 social preview image for a new blog post so shares look consistent on social platforms.
- Generate a blog hero image with a 2K resolution for an editorial article header.
- Produce a clean product photo on a white background for a landing page, then add alt text.
- Make a tall infographic visual for a data-heavy page, using the infographic mode.
- Create a 2:3 Pinterest pin for a recipe or how-to article.
Frequently asked questions
How do I install the seo-image-gen skill?
Download the seo-image-gen ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills. The skill also needs the banana extension, installed with ./extensions/banana/install.sh.
What do I need to run seo-image-gen?
You need the banana extension installed, a nanobanana MCP server connected to Gemini image generation, and a Google Gemini API key from Google AI Studio. The skill's instructions also suggest ImageMagick (the magick command) for converting images to WebP.
Is seo-image-gen free?
The skill is released under the MIT licence. Image generation itself is billed by Google through the Gemini API, and the skill's documentation says each generation costs money. Check Google's current pricing page before you generate in bulk.
What is the difference between seo-image-gen and general image generators?
seo-image-gen is scoped to SEO assets and sets aspect ratios and resolutions for uses such as OG images and Pinterest pins, then adds SEO steps like alt text and schema. It is explicitly not intended for general artwork, banners or thumbnails unrelated to SEO. A general image tool will not apply those SEO defaults or checklist for you.
Why does the skill say it cannot find the image tools?
The skill checks whether the gemini_generate_image or set_aspect_ratio tools are available. If they are missing, the MCP server is probably not connected, so run ./extensions/banana/install.sh or re-check your MCP setup and API key. The source also suggests generating a new key at Google AI Studio if the key is invalid.
What should I do if an image is blocked or I hit a rate limit?
If the output is blocked with IMAGE_SAFETY, rephrase the prompt, since the skill's references include guidance on adapting prompts for safety. If you get a 429 rate-limit error, wait 60 seconds and retry, and check current free-tier limits before running batches.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use seo-image-gen if you run a website or blog and want quick, correctly sized AI images for social previews, heroes, product shots, infographics or Pinterest pins, and you are comfortable setting up a Gemini API key and paying per generation. Pick something else if you need general artwork, strictly controlled costs, or photographically accurate product or people images, since AI output may not match reality.
Download the skill