Test Whether a Shorter CLAUDE.md Helps

Three hundred lines of instructions and it still breaks your conventions. So you add more rules. It gets worse.

My working hypothesis is that a long instruction file may make important rules harder for the model to follow. First check that the file loaded and that its rules do not conflict.

For every line, ask one question: would deleting this cause a mistake? If not, cut it. Delete anything inferable from reading the code. Delete generic advice. Link to API docs instead of pasting them.

Test the change on comparable tasks and record which conventions each run follows. A shorter file is a hypothesis to evaluate, not proof of better adherence.