My big frustration with skills is they're not consistently invoked
My big frustration with skills is they're not consistently invoked. In this post, Vercel saw the same issue.
Their solution? Put a compressed index of docs in AGENTS .md instead.
The improvement in results might be worth it for more general knowledge.
https://vercel.com/blog/agents-md-outperforms-skills-in-our-agent-evals