Over the past few weeks, I have been building AI agent “skills” using Anthropic’s Skills convention.
It is like structured prompting, but upgraded: each skill bundles the context an agent needs and can include small deterministic scripts it can run when needed. The context can be set up in a way that the agent does not upload all at once, avoiding context-bloating and eventually poor performance.
The Skills convention makes agents more reliable and much easier to reuse.
I have been developing these with Claude Code, but the best part is portability: multiple top models can follow the same convention. Skills are an organised, shareable layer across platforms.
Most of what I build is either lyfX-specific or too IP-sensitive to share. But I have now open-sourced some skills that are general enough to be useful for others. I will be adding to this repo over time.
View the open-source skills repository on GitHub
EN 
DE 
PT 
ES