I've fixed a bunch of bugs in my app already, and I don't want any of them coming back. I want regression tests for the bugs I've already squashed. I'll describe bugs I've fixed (or point you at commit messages, a changelog, or issue tracker entries). For each one: 1. Understand what the bug was: what input or sequence triggered it, and what the wrong behavior was. 2. Write a test that would have FAILED against the buggy code and passes against the fixed code. If you can't tell from the code what the buggy version looked like, ask me. 3. Put the test in the right place in my suite with a name that references the bug, so future me knows why it exists. Where the bug was user-facing (a broken flow, a wrong calculation), prefer a test at that level. Where it was internal, a focused unit test is fine. Don't write tests that pass against both the buggy and fixed versions. A regression test that can't catch the regression is decoration. When finished, run the new tests and give me a table: bug, what triggered it, and the test that now guards it.