The pruning engine uses regular expressions to extract all inline markdown links and autolinks, verifying their HTTP status codes in parallel using streamed range-request GET probes. Rather than removing entire sentences, the engine unwraps non-responsive URLs into plain text by replacing the markdown syntax with the anchor text, which preserves natural sentence flow. This process applies replacements in reverse character-offset order to maintain string integrity and strictly protects the client target placement, reporting any target failure rather than pruning it.