Generally, my script analyses code comments and try to find td-like keywords (mainly TODO, workaround, etcs) to create issues based on pieces information like: commit link, author, code snippet and so on. I could apply some ML to analyse such comments, but at this moment I am first trying to validate these initial issues. I do not check for duplicates yet. I can check those indicated in the comment itself (like those marked with any code). Besides, I would have to try make this link with other issues created in the repo (with no indication). I tried to analyse the ones I generate above, they do not seem to be duplicated. Do you with we would create them on the main repo?