I had similar needs - specs being kebab-case, but API is specified and validated for underscore. Didn't find an acceptable solution - we were working off of json schemas and so I ended up using those for in-production validation, had middleware to kebab-case any request data, then used kebab-cased specs internally
not the cleanest setup and hoping to revisit this possibly when spec2 comes out, which has some improvements for programmatically making specs. that could make it easier to alter a spec to use a different casing
Im thinking of doing the same thing, one middleware to change the swagger json to camelCase and another one to convert the data to kebab case when a request arrives