Skills is just a tech bro word for a simple markdown file with instructions.
No need to over complicate it. Write down things you feel like re-using. Like how to specifically implement something in your system ("when adding a new API endpoint we need to do x y and z", or "when making a github PR we tag Æ and Å") so you don't have to repeat it. And I mostly add it in cases where it didn't infer it itself. So very reactive, not proactive.
Most public skills are useless and over complicated. Lots of people are spending too much time on their harness, than actually making stuff.
Edit: but do get inspired by public ones. For instance a "grill me" skill can ve be useful, but I find the public one very mumbo-jumbo. But the idea of forcing the agent to ask clarifying questions is good.
I believe skills are much more than a simple markdown file with instructions. They are a very powerful script engine. How I organize it is that of course there's a markdown with instructions but I split the work in a hybrid of script + instructions. So all the work that can be deterministic is a python script api surface and all the logical or thinking work is in instructions. and agent is also instructed on how to use the api of the python helper functions. This makes it almost like a normal script but the runtime is a harness and the business logic can be any combination of code + human-level intelligence.
So I like to do all the edge case handling and validation etc via a helper function, and the agent is simply instructed to call the function to do something. It is extremely powerful and a completely different way of automating things. I am constantly forced to re-think how computers are supposed to work and its limitations.
My genuine question is:
Are there any "skills" at all that have proven to be useful? And if so, what's the context?
Because, for me anyway, LLMs usually do one thing, and that then produces a durable artifact. So the prompt that got me there by that point expired and is not really needed anymore.
I also occasionally have recurring tasks (rarely though), but there, the prompt to do stuff is embedded in code that orchestrates the doing, so I have no use-case for that either.
___
For the "add this endpoint" example you've described, I just throw commit IDs at the clanker and say "go do that again". That works, and doesn't decouple knowledge from code.
[dead]