While we're on the topic, I think styling DSL would be much more readable with some minor adjustments Consider this:

fun RenderContext.headerImage(srcUrl: String, alternativeText: String) {
    (::img.styled {
        boxShadow { flat }
        radius { large }
        width(sm = { small }, md = { smaller })
    }) {
        src(srcUrl)
        alt(alternativeText)
    }
}

VS reimagined:

fun RenderContext.headerImage(srcUrl: String, alternativeText: String) {
    img.styled({
        boxShadow { flat }
        radius { large }
        width(sm = { small }, md = { smaller })
    }) {
        src(srcUrl)
        alt(alternativeText)
    }
}

or even forget inline styles and just have optional style arguments in existing DSLs:

fun RenderContext.headerImage(srcUrl: String, alternativeText: String) {
    img(style = {
        boxShadow { flat }
        radius { large }
        width(sm = { small }, md = { smaller })
    }) {
        src(srcUrl)
        alt(alternativeText)
    }
}

// OR with builder to allow reusable styles as well
fun RenderContext.headerImage(srcUrl: String, alternativeText: String) {
    img(style = styleOf({
        boxShadow { flat }
        radius { large }
        width(sm = { small }, md = { smaller })
    })) {
        src(srcUrl)
        alt(alternativeText)
    }
}