improve-testslisted

# Improve Tests Measure a test suite, find the tests that cost time and catch nothing, find the setup that wastes time, and cut both. The goal is the shortest run that still catches the bugs this project actually makes. Starts from a measured baseline and an assessment: if the suite is already lean, say so and stop. Test files are the target, so editing and deleting them is expected. Production code changes only where a test needs a seam (Step 7), and only as the approved plan names it. If a cheap test is impossible because module boundaries are wrong across several modules, that is architecture work: say so, and suggest the `improve-architecture` skill if it is installed. ## Principles Every step applies these. Findings cite them by name. - **Name the break.** A test earns its place by the break it catches: a realistic change to production code that turns it red. A test with no nameable break costs time and catches nothing. Test as little as reaches the confidence the project needs, and spend tests where the project goes wrong: complicated conditionals, money, security, code with a bug history. - **Our data, not their library.** Where our code hands work to a library (PDF, email, image, chart, spreadsheet, HTTP client, ORM, framework router), the library owns the rendering and has its own tests. Our claim is the data we hand it. Build that data in a plain function or object (the document model, the template variables, the request payload) and assert on it in a fast test