We certainly need to know something about the agent's capabilities. Humans certainly aren't capable of fortelling the future and knowing exactly what adaptations will be beneficial in the future. If Humans were attempting to engineer systems as flexible and as adaptive a living things, they would turn to genetic algorithms.
Where did the genetic algorithms come from, intelligent design or sheer happenstance?
Nope. When looking at the evidence of the human insulin gene, we can justifiably hypothesize an agent without regard to his capabilities. What you're describing is an attempt to define the characteristics of the agent, based on the fact that certain evidence points to his existence -- which is a much different problem then trying to determine how a bacterium came to be producing human insulin.
Humans certainly aren't capable of fortelling the future and knowing exactly what adaptations will be beneficial in the future.
That's clearly an improper characterization of genetic engineering as it's practiced today -- which is primarily driven by achieving specific, beneficial, and future goals. Your complaint seems to be that the current state of technology is geared toward "the next step," and does not exlpicitly include longer-term goals. But that's simply the current state of technology and commerce; there is no reason to assume that advances in the field will not include longer-term predictability in the future?
If Humans were attempting to engineer systems as flexible and as adaptive a living things, they would turn to genetic algorithms.
Well, OK -- so you've answered your own complaint, haven't you? Humans are already attempting to engineer living things from scratch, based on an algorithmic approach; moreover, the Human Genome project seems to be yielding something of an inverse effect, in that decoding the genome apparently shows the presence of "genetic algorithms," including levels over and above the strict DNA sequence. In either case, an "algorithmic" approach to genetics seems within reach (albeit perhaps not particularly soon).