There's a term for what I've been focused on: AX. Agent Experience.
It's been around a while. DX for the agent in your codebase. How the code is structured. How fast the feedback loops are. Whether the information an agent needs is easy to find.
That's where most of my time has gone recently. Tuning CLAUDE.md, tightening READMEs, refactoring naming that confuses agents, adding hooks and skills that close feedback loops. It's why I wrote your codebase is your prompt. Everything in the repo is context. The agent reads all of it.
Same levers as DX. Sharper requirements. Agents can't shrug off a stale README. They take the loudest signal in the room.
If the code is unclear to you, it's unclear to the agent.