Claude Code is strongest when the definition of done is executable. Test-driven development gives the agent a tight loop: reproduce the bug, make the test pass, then refactor only if needed.

A Good TDD Prompt

Bug:
Expired invite tokens are accepted.

Task:
Write the failing regression test first in the closest test file. Then implement
the smallest production change that makes it pass.

Verification:
Run bun test:unit -- src/lib/server/invitations.test.ts and bun run lint.

Stop:
If either command fails for an unrelated reason, stop and report the failure
instead of editing unrelated files.

This prompt gives Claude a behavior to reproduce, a scope boundary, and a finish line.

Watch the First Test

Do not let the agent write a test that already passes. Ask it to report the failing assertion before implementation:

After writing the regression test, run it and show the failing assertion before
changing production code.

That one sentence prevents a lot of fake coverage.

Keep Refactors Separate

After the test passes, ask whether a refactor is necessary. Many fixes are better left small. If the implementation is clear and the tests pass, stop.

When a refactor is justified, run the same verification command afterward. A green test before a refactor does not prove the refactor is safe.