Two skills. One makes AI answer first and stop. The other makes it sound like a person. Both ready to copy.
There are two different AI writing problems, and almost everyone tries to fix them with one prompt.
When AI writes for other people (an ad, an email, a landing page), the problem is that it sounds like AI and your reader can tell.
When AI talks to you, the problem is volume. You asked one question and got six paragraphs, and four of them restate the question before answering it.
Those need two different skills, so here are both. The short one is built on three ideas worth knowing.
01
Simplified Technical English
A controlled writing standard (ASD-STE100) first built for aircraft maintenance manuals, where a misread sentence is a real problem.
The current edition has 53 writing rules and a dictionary of about 900 approved words, each with one meaning. Active voice, short sentences, one instruction per sentence. You don’t need all 53 rules. You need the habit of one meaning per word. (About STE)
02
Attention Span
A free, open-source set of output styles for Claude Code that changes how it talks to you, not how it codes. It gave me the answer-first rule. Put the conclusion in line one and say the least that fully answers, then stop.
Its own benchmark found output about 43% shorter on average, with the test pass rate at 97% both with and without the style. Shorter didn’t mean worse. (The repo)
03
Zinsser’s articles of faith
From On Writing Well, the nonfiction classic first published in 1976. William Zinsser built it on clarity, simplicity, brevity and humanity.
Brevity does the most work here. His line is "Examine every word you put on paper. You’ll find a surprising number that don’t serve any purpose." Humanity is why the skill below doesn’t turn your AI into a robot that answers in fragments.
Use this when AI is talking to you. Paste it into your custom instructions or project instructions, or save it as a skill.
Use this when AI writes something other people will read. The ban lists are the real payload. They’re the exact words and sentence shapes that make a reader think "this was AI" before they can say why.
These two do opposite jobs, so don’t run them in the same place.
SKILL.md inside its own folder: ~/.claude/skills/get-to-the-point/ and ~/.claude/skills/humanizer/. In Codex, the same folders go in ~/.agents/skills/.One honest note on cost. Fewer output tokens saves real money if you pay per token through an API. On a flat monthly plan it mostly saves you time and attention, which is the better reason anyway.