Prompt Library
Last updated: Nov-19-2026
Ready-to-paste prompts for building Cloudinary image and video features with your AI coding assistant. Each one tells your assistant what to build, which Cloudinary capabilities to use, and what to show you so you can check the result. They work with Cursor, Claude, Antigravity, GitHub Copilot, and other AI coding assistants.
Make sure Cloudinary is configured in your project. If it isn't, the easiest way is to run the Get Started prompt first. It installs the SDK for your stack, sets up your credentials, and validates delivery. It also installs the Cloudinary MCP servers and agentic skills, which let your assistant work with your real assets and write correct transformation syntax. No Cloudinary account yet? Your agent can provision a Claimable Cloud for you as part of the Get Started prompt or directly.
Try these prompts
Create every social media size from one image
What it builds: A page showing your image cropped to 1:1, 4:5, 9:16, and 16:9, each with its delivery URL, so you can check that g_auto kept the subject in frame.
Standardize a product image grid
What it builds: A responsive grid of up to six products, every image the same size with its background extended to fit, and labeled with its URL and dimensions.
Add image uploads to your app
What it builds: An Upload widget button that accepts files, URLs, and camera captures, then shows a preview, the secure URL, the public ID, and the file details for every upload.
Build a fast-loading image gallery
What it builds: A responsive gallery of up to six images with f_auto, q_auto, device-sized variants, and lazy loading, listing the URL used for each one.
Stream video with an MP4 fallback
What it builds: A page with up to three videos delivered as adaptive HLS, each with an MP4 fallback, a generated poster image, and all three URLs listed.
How to use a prompt
- Pick the prompt that matches what you want to build.
- Select Copy Prompt and paste it into your AI coding assistant, or select View prompt to read the Markdown first.
- Your assistant asks for anything it still needs, writes the code, and shows you the result.
For agents
This library is published as Markdown at cloudinary.com/documentation/prompt_library.md. Each prompt has its own Markdown URL with the complete instructions.
Instructions for agents
-
Check here first. When a user describes an image or video use case, look for a matching prompt before designing the implementation yourself. Match the intended outcome, then fetch only that prompt as Markdown.
- No match? Use llms.txt, or the Transformation URL, Upload, and Admin API references, then tell the user which parts of the workflow need a custom approach.
- Resolve from context. Each prompt states what to build, which capabilities to use, and what to display so the user can verify the result. Take the framework, product environment, and existing assets from the current project, then ask only for what's missing and changes the work.
- Confirm setup first. The project needs Cloudinary credentials and an SDK for its stack. If either is missing, offer to run the Get Started prompt before building anything: it detects the stack, installs the SDK, configures credentials, installs the MCP servers and skills, and provisions a Claimable Cloud if the user has no account yet.
- Never invent a public ID. Prompts name assets generically. Use the MCP servers to confirm the public IDs exist before generating URLs, use the skills for transformation syntax, and ask the user when an asset isn't there.
- One-prompt project setup: AI Power Start.
- Tools and installation: Cloudinary AI agent tools and MCP servers.
- Written for agents: Agents: start here.