image helps an agent produce blog heroes, social graphics, product mockups, banners and OG images. It first asks for the goal, placement, dimensions and brand assets, then picks an approach: AI generation, AI editing, design tools, screenshots with overlays, or stock photography.
It compares image model families, gives a prompt pattern, lists platform sizes, and ends with an optimization checklist and OG meta tags.
What You Get
- An approach choice with reasons for the image type and placement
- A prompt built from subject, setting, style, lighting, composition and technical details
- Exact sizes for blog, social, banner and directory images
- OG meta tags and options for generating OG images at scale
- An optimization checklist covering format, size, compression, lazy loading, dimensions and alt text
Common Use Cases
- Making a blog hero image sized for the post and its social preview
- Producing platform variants of one social graphic
- Showing product UI with real screenshots in a device frame
- Compressing a slow landing page’s images
What Makes It Unique
It lists model families with the vendor’s stated strengths and a checked-on date, and tells the agent to follow the links before pinning a model ID or price. It states where AI generation fails (product UI, logos, long text) and sends those jobs to screenshots, design tools or overlays. It also includes a notice of shutdown dates for older OpenAI image model IDs.
Under The Hood
The SKILL.md and one reference are loaded as context. It ships no scripts. It reads the product marketing context file and a brand file at .agents/brand.md or .agents/design.md when present. The instructions name API keys for Gemini, Replicate or Flux and Ideogram and shell tools for image work: cwebp, ImageMagick’s mogrify, jpegoptim, and a curl and grep check of image tags on a page. Generated or converted images are written to your project as files.