Anthropic Just Made Every AI Framework Optional.

In October 2025, Anthropic shipped this. Most teachers still have not caught up. They called it Agent Skills, a new open standard for talking to Claude. One file, plain text, file name skill.md. Three lines at the top, plain English body. No imports, no SDK, no new programming language. The smallest, simplest unit of AI agent definition ever shipped by a major lab. Simon Willison, the engineer who built Django and runs one of the most read engineering blogs on the planet, wrote this the same day. Word for word, Claude skills are awesome, maybe a bigger deal than MCP. If you do not know what MCP is, that is fine. The point is, the people paying attention saw something most teachers missed. So, why did this matter? Because every AI agent framework on the market, LangChain, AutoGen, CrewAI, all of them, was solving a problem that Claude itself was about to solve natively. When the model itself handles the loop, the tool calls, and the orchestration, you do not need a framework to bolt those on top. You need a file telling Claude what you want. Here is what is inside a skill.md name, description, when to use it, then plain English instructions. Read these files, find what changed, write a status, save it. No code, just words. That is the spec. And because it is an open standard, the file you write today runs on Claude code in your terminal, Claude on the web, Claude on your desktop, and any other tool that adopts the standard later. One file, multiple harnesses, no vendor lock-in. This Saturday, inside Agent Skills Academy, I will walk you through writing your first one. Free. 75 non-coders are already shipping theirs every weekend. The standard is here. The frameworks will fade.