What it does: The other four checks measure a draft the way a database would. None of them notice the thing a reader notices in two sentences — that nobody wrote this. The lint reads the body and returns findings, not a score: each one carries the check it came from, its severity, the sentence quoted, the fragment matched, and an explanation in words. How to access: Draft Forge > Settings > strictness; the findings appear under the queue row’s check summary and on the refresh review screen. Free/PRO: Free, entire. It reads a draft the credits already paid for; nothing in the file asks about a licence, and running it costs nothing. What it looks at — six families:

• Stock phrases. A bundled list of 132 English clichés, measured as a density per 1000 words rather than counted, so a long piece is not punished for one of them.
• Repetition, four ways: the same sentence opening three times or more, sentences repeated verbatim, the proportion of four-word sequences that recur, and paragraphs built to the same shape over and over.
• An empty introduction: the opening restating the title with no concrete detail, or promising what “this article will” do.
• Readability: Flesch reading ease, mean sentence length, and a rhythm too even to have been written by a person.
• Numbers and dates: a year before 1900 or more than a year ahead, a percentage above 100, “studies show” with nobody named in the same paragraph.
• Structure: headings against the volume of text, and a list standing in for prose.

Notes:

• Strictness is only about what blocks. All the checks always run at every setting. The three presets — lenient, standard (the default), strict — differ in their numbers and in which findings they are willing to stop a draft over. A blocking finding closes the gate; a warning rides in the report and stops nothing.
• Below 120 words the prose checks say nothing: there is too little there to judge.
• The language is named where it matters. The phrase list and the Flesch score are English and say so. Everything else — sentences, paragraphs, headings, repeats — a Russian or Japanese draft has exactly as much of. A non-English body comes back with fewer findings, never with nonsense ones.
• The phrase list ages. It can be replaced wholesale by a stored list (option draftforge_slop_phrases), so a later Forge API feed can refresh it without a plugin release, and overruled per-site by the draftforge/slop_phrases filter.
• The list is a snapshot of what models were writing when it was compiled, and the per-1000-words threshold was calibrated on fixtures rather than on a sample of real drafts. That is the honest limit of it today.