The question is in the title. It’s currently unclear when we should use skills over rules and vice versa. Are skills a replacement for Cursor rules? If not, why?
1 Like
Rules are working
Skills are not
Your welcome 
3 Likes
I’m worried that skills and rules in the Cursor harness will just waste tokens for nothing, especially with sub-agents, skills, and other added features.
We also don’t currently know the context size of the Cursor harness. I think the Cursor team should allow users to opt out of features that aren’t really useful.
(For example i opted out from the browser tool)
1 Like
You can see this in the chat indicator. My User Rules are ~1.4k tokens + ~2k tokens usually added as a system prompt and context from indexing. For a context window containing hundreds of thousands of tokens, this is a grain of very important information.
99.99% code and ~90% documentation in my projects are written by agents, and I don’t see or notice any unnecessary Cursor features. I just don’t use Cloud Agents, but that’s because I’m afraid of wasting money and losing control, not because they’re useless.
It’s useful for searching documentation and references.
To be clear, this is specific to subagents, who don’t automatically learn skills.
@Colin Sorry, it’s unclear whether the “Apply intelligently” cursor rule should be migrated to skills so the model will use it more effectively ?
Skills are always applied intelligently, so no need to put this in the frontmatter (it won’t do anything).
Also, if you haven’t seen it, there’s a command to migrate skills to rules.
@Colin What is the real benefit of migrating? Has the Cursor harness changed, making the use of skills more appropriate than intelligently applied Cursor rules? Or do skills behave exactly the same and mainly exist for interoperability across AI IDEs and CLIs?