Claude Fable 5.1's writing is generally preferred to earlier models', but it has habits worth knowing about: denser prose than before, less formatting where you might want some, and — deep into long sessions — a tendency to write for someone who watched it work rather than someone who didn't.

Lead with the outcome

The most useful communication instruction, and it works on every task:

lead-with-outcome.txt
Lead with the outcome. Your first sentence after finishing should
answer "what happened" or "what did you find" — the thing I'd ask for
if I said "just give me the TLDR." Supporting detail comes after.

Being readable and being concise are different things, and readability
matters more. Keep output short by being selective about what you
include, not by compressing the writing into fragments, abbreviations,
or arrow chains.

After a long run, write for a stranger

When it has been working for a while — many tool calls, a big context — its final message can be full of shorthand it built up while working: names it made up, references to decisions you never saw, arrow chains, packed identifiers. This addendum fixes it:

regrounding-summary.txt
Terse shorthand is fine between tool calls. Your final summary is
different: it's for a reader who didn't see any of that. Write it as a
re-grounding, not a continuation of your working thread.

Open with the outcome in one sentence. Then the one or two things you
need from me, each explained as if new. Drop the working shorthand:
complete sentences, terms spelled out, no arrow chains or labels you
invented earlier. When you mention a file, commit or flag, give it its
own plain-language clause saying what it is or what changed — never
several packed into one parenthesised run.

If you have to choose between short and clear, choose clear.

Density

Fable 5.1's prose can run to longer sentences and fewer paragraph breaks than earlier models. If it reads as heavy, ask for plainness — and put the instruction in your first message rather than the system prompt, where it holds better:

plain-prose.txt
Please remove all mannered prose. Say what you mean. When a literal
phrase is available, use it — no metaphor standing in for a direct
statement.

Formatting

Where earlier models overused bullets and bold, Fable 5.1 does the opposite: fewer headers, fewer lists, less emphasis. If your existing prompts contain anti-formatting rules — "no bullet points", "don't use headers" — remove them; they now push in the wrong direction. If you want structure, ask for it conditionally:

formatting-rule.txt
Use lists and bullet points when asked to, or when the content is
multifaceted enough that they help. Keep conversational exchanges to
plain prose.

When summarising documents

It's more likely than earlier models to reproduce passages from a source without marking them as quotations. If you're building a summariser, the fix is one complete example in your system prompt of a correct response — the request, the answer written in the model's own words with at most one short marked quotation, and a sentence saying why that's correct. One good example beats a paragraph of rules.

Put it somewhere permanent

None of this needs repeating per message. In Claude Code, it belongs in CLAUDE.md. On the API, in the system prompt. In a chat interface, in your first message of a session — style instructions placed there hold well for the whole conversation.

If the summary needs you to have watched the work to understand it, it isn't a summary. Ask for it to be rewritten for someone arriving cold.