en-checklisted

# English Text Quality Check Review the text in $ARGUMENTS. With no arguments, review the most recent English text in the conversation. If $ARGUMENTS names a file or a glob, read those files. ## Where the corpus lives This skill reads the reference files that ship with the **en-text** skill, installed alongside it. Locate the folder once, then read from it: - Find a directory named `references` whose parent directory is named `en-text`. It sits next to this skill, typically at `../en-text/references/`. - Read `clarity.md`, `slop.md`, `usage.md`, and `method.md`. Read `scoring.md` only if the user also asked for a number. If you cannot find the corpus, say so and stop. Do not improvise rules from memory and present them as en-text findings. ## Procedure **1. Frame the text.** Before reading for errors, settle five things and state them in one line: genre, audience, spelling variant (`U28`), whether a house style applies (go look for one, `M3`), and whether the text is source or a rendering of it (`M2`). Every later judgment depends on these. A landing page and an RFC do not get the same review, and a fetched page cannot answer the questions a dictionary file can. If the request names recent edits and asks whether they broke anything, settle one more thing now: do you have a before state? If not, say so in the same line, and report findings without attributing them (`M1`). **2. Triage pass.** Read the whole text once without annotating. Note only: - word count of