en-textlisted

# en-text — English Text Quality An independent corpus of English writing rules, written for AI agents rather than for people. Every rule has an ID, a test you can run on a sentence, and a fix. Findings cite the ID so a reader can argue with the rule instead of with taste. Credits and recommended reading: `references/sources.md` ## What this skill is for Two problems stack on top of each other. Bad prose existed long before language models: buried subjects, nominalizations, hedging, sentences that make the reader hold six things in memory before the verb arrives. Then models added a dialect of their own: *delve*, *leverage*, *it's not X, it's Y*, three items in every list, a bolded bullet for every thought. A text can be free of AI tells and still be unreadable. A text can be perfectly clear and still announce on line one that a machine wrote it. This corpus treats both. ## Priority order When rules conflict, resolve in this order: 1. **The user's explicit instruction.** A requested register (legal, academic, marketing, deliberately baroque) overrides any default here. These are defaults, not mandates. 2. **Accuracy.** Never trade a true statement for a smoother one. If tightening a sentence would change what it claims, keep the claim and say the sentence resists compression. 3. **The reader's effort.** Between two accurate versions, the one that costs less to read wins. 4. **Everything else in this corpus.** ## Three hard boundaries **Reviewing is not