Guides

How to make Claude answer shorter: 16 rules borrowed from aircraft manuals

Ask Claude a yes-or-no question, and you can get an essay. The fix is a short set of rules in the instructions Claude reads before every chat. Ours come from Simplified Technical English, the standard that aircraft maintenance manuals are written in: the answer in the first line, one thought per sentence, warnings before the action, and the decision Claude needs from you on the last line. Below are the 16 rules to copy, where to put them in the Claude app and in Claude Code, a one-line version for a single question, and when to let Claude write long.

The 16 rules (copy them)

Paste them as they are. They are written the way they ask Claude to write.

The answer
1. Answer in the first line. Details after.
2. Don't retell how you worked. Only the result, the risks and what you need from me.
3. Put risks and warnings before the action they apply to.
4. The decision you need from me goes on the last line.
5. Give details only when I ask.

Sentences
6. Use the active voice.
7. One thought per sentence.
8. Max 20 words in an instruction, 25 in a description.
9. One topic per paragraph, max 6 sentences.

Steps
10. Write instructions as commands: "Check X."
11. One action per sentence. Steps as a numbered list.
12. Condition before the action: "If X, do Y."

Words
13. Always call the same thing by the same word.
14. No idioms, no jargon.
15. No more than three nouns in a row.
16. Exact numbers, not estimates: "in 2 hours", not "soon".

Why aircraft manuals

In the 1980s, European airlines asked the aerospace industry for a form of English that mechanics could follow without misreading, including mechanics who aren’t native speakers. The result became ASD-STE100, Simplified Technical English. The current issue, from January 2025, has 53 writing rules and a dictionary of about 900 approved words, and it’s free to download.

Some of its rules: use the active voice; no more than 20 words in an instruction and 25 in a description; one instruction per sentence; one topic per paragraph and no more than six sentences in it; one word for one meaning; and no more than three nouns in a row.

Reading Claude between tasks is a similar job. You scan, you act, you go back to work. Our 16 rules aren’t the standard: they borrow its sentence and word rules (6–16) and add five rules about the shape of an answer (1–5).

Where to put them

Put the rules where Claude reads them before every message, so you never paste them again.

In the Claude app (web, desktop, phone)

Click your initials in the lower left corner, then Settings → Your Claude → Instructions, and paste the rules. In the older settings layout it’s Account → Instructions for Claude. Anthropic calls these profile instructions: they apply to all your conversations.

In a project

If you want short answers in one project only, put the rules in that project’s instructions instead. They apply only to chats inside the project. Projects are on paid plans.

In Claude Code

---
description: The answer first, the decision last
keep-coding-instructions: true
---

The 16 rules go under the header. The file name becomes the style name.

Through the API

Put the rules in the system prompt.

None of these is a hard limit. Claude reads them as instructions, and the docs say so plainly: an output style “doesn’t guarantee that something always happens or never happens.” If an answer slips back into an essay, reply with the number: “Rule 1.”

For one question only

No settings, one chat. Add this line to the question:

Answer in the first line: yes or no. Then at most two sentences on why.

When to let it write long

Short isn’t always right. Ask for the long version when:

Rule 5 covers all three: write “details”, or “ignore the length rules for this one”.

Where Meowdoro fits

We built Meowdoro, a focus timer around the MacBook notch for people who work with Claude Code and Codex. When a session needs you, the notch shows it with the question it’s waiting on. With these rules, that question is the last line of a short answer: you read it, you answer, and you go back to your focus block or your break. Try it and tell us what’s missing.

Questions

Why does Claude write such long answers?

It doesn’t know how much you want, so without instructions it covers the question from several sides. Some models also start out wordier: Anthropic’s prompting guide says Claude Opus 5’s default answers run longer than earlier models’, and that lowering effort doesn’t reliably shorten them. Its advice is to ask for brevity in the prompt. Do it once, in the instructions Claude reads before every chat.

Is there a setting for shorter answers?

In Claude Code, yes: the built-in Concise output style (/output-style concise). In the Claude app, put the rules in Settings → Your Claude → Instructions, and they apply to every chat.

Will shorter answers leave out something important?

Rule 2 keeps the risks and the question for you, and rule 3 puts warnings before the action. What goes is the retelling of how Claude got there.

Do shorter answers use less of my usage?

A little. Everything Claude writes counts as output, and output counts toward your limits. The bigger saving is your time.

Where instructions go and how output styles work: the Claude Help Center on personalization, and the Claude Code documentation on output styles and CLAUDE.md. Answer length in Claude Opus 5: Anthropic’s prompting guide. All read on November 19, 2026. Simplified Technical English: the ASD-STE100 site and Wikipedia; the three-noun limit as listed by Boeing’s Simplified English Checker. The 16 rules are ours.