llm-platform-security / chatgpt-plugin-eval Goto Github PK

Dear Authors,

Thanks for your research! I'm interested in your study.
And I wonder how you crawl OpenAI plugins. Specifically, in the paper you mention this:

"We visited the OpenAI plugin store and individual plugin developer websites to download plugin manifest and specifications. We downloaded the amalgamated manifests for all plugins from the OpenAI’s official plugin store. We then programmatically traversed the plugin manifests and sent requests to each plugin services’ API URL to download their API specifications. Additionally, we also download privacy policies of plugins from the links provided by plugins."

Upon visiting the plugin stores, I was only able to find links to the privacy policies provided by the plugins, but I did not come across any manifests. Could you kindly clarify how you were able to obtain these manifests and specifications?
Additionally, since you mentioned that you "programmatically traversed the plugin manifests...", would it be possible for you to share the corresponding script?

Thank you once again for your valuable research.