A short talk from AI Engineer conference in which two Anthropic engineers (Barry Zhang and Mahesh Murag) make the case that you don’t need to build agents. Instead use a general purpose agent (Claude Code) and then write skills (skills are just pe-canned prompts expressed as markdown). The advantage of this approach being that skills are simple, versionable and composable. This last point seems the most important, no more wrangling graphs of actions.

MCP is not fully deprecated in this approach, skills sit as a layer in front of MCP pointing back to the endpoints available to the agent to get the task done. A common criticism of MCP servers has been how quickly they fill the context window. So adding this layer and then fine tuning the model to learn how to self discover skills makes a lot of sense.