Is it possible for cursor to "self-upload" images?

I am working on an agent that can generate/fix shaders (specifically ISF shaders). AS part of that, I wrote a command line tool that takes a shader file, compiles, runs, and renders frames of the shader to image files.

I would like the agent to be able to take the outputted image files and automatically upload into the chat context, allowing it to analyze the visual results of the rendering. It can then use that analysis to ensure that the end result matches the intention of the code it generated.

Is it possible for the agent to self-upload an image file without my intervention?

agent has ability to search a folder of files if asked to, you could tell it after the rendering is done to search the output folder for latest files and analyze them, otherwise no, it cannot be uploaded to context automatically, to my knowledge

I tried that, and it will look at the file sizes or even write a program to analyze the files to look for empty/black files. However, it really improves the original code when I upload the image for analysis.

Is it an option to ask it to open the content in a browser and use what it sees there? Use playwright mcp to get it to launch the browser. GitHub - microsoft/playwright-mcp: Playwright MCP server

you could possibly achieve a better result with mcp but you might need to write it yourself