"Write Clean Code" Is a Hope, Not an Instruction
AI coding agents remix what they already find, and most real codebases are mixed. Telling one to "follow SOLID" or "write clean code" sets no standard — until you show it good and bad, and make it pro
Use TDD. Refactor carefully. Follow SOLID. Use design patterns. Write clean code. I've seen agent instruction files full of phrases like this. The problem isn't that any of it is wrong. It's that none of it is self-executing.
An AI coding agent mostly remixes what it already finds in the codebase. If the surrounding code has clear boundaries, strong tests, consistent naming, and good examples, you have a real chance of useful output. But most real codebases are mixed, some good patterns, some old shortcuts, some accidental complexity, a few "don't touch that" corners, and code that made sense three years ago and now quietly works against the team. Drop "write clean code" into that and you are not setting a standard. You are expressing a hope.
The concrete version works better. Show the agent what good looks like and what bad looks like. Show it which patterns matter in this repo and which to avoid. Then the part most teams skip, make it prove what it actually followed.
The reviewer should not have to guess whether the agent met the team's engineering standards. The agent should show what it looked at, what it applied, where it applied it, and how it verified the result. That is the difference between putting best practices in an instruction file and building them into the delivery workflow.
"Write clean code" is a hope. "Show me which pattern you followed, and prove it held" is an operating model.