AAS Core's skill picker for AI agents hits 43,574 stars

AAS Core's Python control plane now recommends which AI agent skills to load, per its GitHub listing. That solves a problem I've hit myself: past thirty skill files, there's no principled way to decide what belongs in context. My own workaround has been loading everything and hoping the agent sorts it out, which works exactly as well as it sounds.

The repo passed 43,574 stars, and it isn't another curated list of links. It ships a CLI and a local MCP server instead. Together they recommend and validate a skill stack for whatever task the agent is running. The catalog behind it holds more than 1,965 entries, per the project's GitHub page.

The GitHub page doesn't say how the recommendation engine scores relevance, or what happens when two skills in a stack conflict. Those are the questions that matter once this runs against a real project instead of a demo.

Skill sprawl is the next context window problem. Most agent setups already carry more skill files than a person can sort through by hand. A control plane that decides for you does a different job than an awesome-list, which only helps you find things.