Show HN: Agent – Inline code annotations for AI agents(github.com)
github.com
Show HN: Agent – Inline code annotations for AI agents
https://github.com/codebasedlearning/agent-annotations-spec
Problem: AI agents miss cross-project relationships — an enum that must stay in sync across repos, a client-side check that isn't the real enforcement gate. Nothing in the code makes those connections visible. Idea: lightweight comments explicit for AI agents.
2 comments
Do you have a few examples of repos where you used this annotation?
I expect that a useful case is to link to the repo that defines the data and the agent can't edit, and you can't add the annotation there. (Like the Linux source tree by Linus himself.) Is it possible to use these annotation when one repo does not collaborate?
I expect that a useful case is to link to the repo that defines the data and the agent can't edit, and you can't add the annotation there. (Like the Linux source tree by Linus himself.) Is it possible to use these annotation when one repo does not collaborate?
Thank you for your comment. To answer your question, I don’t currently have a public repository where you can see the use cases in action. I only have private product (confidential) repositories that I actually implement the annotations in, and ones where I can see added value. The idea is still new. I am gaining experience in order to determine where and whether it would be beneficial.
Secondly, the specification includes the option to incorporate cross-references (agentref), which would enable you to store references to non-modifiable external repositories. Maybe this could do the trick. With these references, the point is to figure out the detail level at which the ref should point to (repo, folder, file, class) such that it may help the agent.