In #281 I'm parsing the HTML string from context.response.body, then transform it, then write it back to context.response.body. Great.
For #153 I would benefit from having access to the parsed response HTML, because I could then walk over the AST, find all a elements, and add their hrefs to the link header to enable pre-fetching.
In #122 I would benefit from having access to the parsed response HTML, because I could then walk over the AST, find all meta elements, and add some of them to the response headers.
Therefore, I propose the creation of a new parse-response-html step, a context.response.hast property and a new stringif-response-html step for the HTML pipeline.