I think a lot of engineers would agree with you that the evidence for TDD, Agile and $fad_framework is weak. Those sorts of debates are a HN staple, so I wouldn't want to criticize the whole field of software engineering because of the popularity of these concepts. Indeed Agile is often criticized for being insisted on by non-developers.

Social sciences aren't criticized for being crackpot anyway, that's a strawman, it's more for the standard reasons that they produce a lot of claims but the evidential basis is weak. If social science was riven with internal debate in which these weak claims were constantly and publicly attacked, then we could compare it to the software world, but in practice we don't see that and dubious claims that come from one or two people often persist for decades before anyone decides to take a second look.