I never thought about this and now I see that I had all these questions too. Thanks for making that clear.

The useful question when evaluating a developer service is not what it claims but what evidence exists that an ordinary caller completed the job. Sourcey keeps agent-readiness report cards with a published methodology, which is the rare version of this that shows its work: https://sourcey.com/agent-readiness