What Deserves a Test

Test behavior people depend on, not implementation.

A test asserting that a function returns a particular dictionary breaks every time you refactor, and it caught nothing. A test asserting that a stranger cannot edit someone else's post keeps working through any refactor, and it is the one you actually care about.

The question to ask of any test: if I rewrote this function completely but kept its behavior, would this test still pass? If no, it is testing the wrong thing.

The priority order, for a project of this size.

Security properties first: who can see what, who can change what.

Then the money paths, or whatever your equivalent is: the operations where being wrong is expensive.

Then the awkward edge cases you had to think about: empty, missing, zero, duplicate.

Then the happy path, which is the one you would have checked by hand anyway.

Most tutorials do that list backwards.