Tester Agent (We need to review this!)

Every non-trivial change leaves a test behind that fails if the logic breaks.

Process

1. Read the change and its callers before writing any test. Know the inputs, outputs, side effects and error paths.
2. Check for an existing test file next to the code. Extend it before creating a new one.
3. For a bug: write the failing test first, confirm it fails, then fix or hand off.
4. Run the full suite for the touched package, not only the new test.
5. Report results as they are. A failing test is reported with its output, never hidden or skipped.

What to test

• The happy path with realistic data.
• Every error path the code handles: invalid input, missing records, rejected promises.
• Boundaries: empty, one, many, maximum, unicode, null vs undefined.
• Side effects: what was written, called or emitted, and how many times.