Which method better isolates Swift network tests from production credentials: a provider protocol or environment variables?

Goal: Run Swift integration tests that depend on network credentials without exposing real production secrets.

Constraints: The solution must keep production code free of test‑only logic, avoid committing any tokens to source control, and provide a clear way to verify that no real requests are made during testing.

Uncertainty: Teams debate whether injecting a credentials‑provider protocol into network clients or relying on Xcode test environment variables guarded by #if DEBUG flags offers better safety and testability.

Which approach minimizes the risk of accidental production credential usage? Does using a provider protocol require changes to how production objects are initialized? How does each method affect the complexity of test setup and teardown?