These are 24 prompts worth saving for writing with Kimi K2.7, the free Moonshot AI assistant at kimi.com. Kimi is a 1-trillion-parameter mixture-of-experts model (~32B active) with a 256K-token context window, and that huge window is why it is so good at long-form: you can paste a full brief, a long transcript, or an entire draft and it still holds the thread. These prompts cover the whole workflow: first drafts, editing, tone, long-form, emails, social, and marketing copy. Every prompt is written out in full, ready to copy, with [BRACKETS] for your own details.

The golden rule with Kimi is simple: give it your real material. Paste your notes, brand voice, examples, or the actual draft, then name the audience, tone, length, and exact output format. It is a strong writer with rich context and a weak guesser from a thin prompt. Leave Instant mode on for quick drafts, turn on Thinking to structure long or argument-heavy pieces, and switch on Web search when a piece needs a real, cited fact. For the full set, see the best Kimi prompts roundup; for reusable starting points, the prompt templates pair well with these.

Advertisement

First drafts

Drafting is where the golden rule matters most. Feed Kimi your notes, an outline, or a transcript — paste it or use file upload — then name the audience, length, and format. It handles the structure and prose so you keep the facts and the point.

1. Blog post from your notes

You are a writer for [AUDIENCE]. Turn these notes into a [900]-word blog post titled "[TITLE]": [PASTE NOTES]. Keep the tone practical and direct, lead each section with the takeaway, use short paragraphs and the concrete examples from my notes, and don't invent facts I didn't give you. Output clean Markdown with H2 sections and a two-line intro.

Why it works: Feeding Kimi your real notes keeps the facts yours; it only handles structure and prose, which is exactly what it does well.

2. First draft from an outline

Write a full first draft from this outline: [PASTE OUTLINE]. Audience: [AUDIENCE]. Target length: [LENGTH]. Tone: [confident but plain]. Follow the outline order exactly, expand each point into a short paragraph with a concrete example, and don't add sections I didn't list. End with a one-line closing thought. Output in Markdown.

Why it works: Turning a skeleton outline into a real draft beats a blank page, and the "don't add sections" line keeps Kimi on your plan.

3. Article from a transcript

Here's the transcript of a [talk / interview / call]: [PASTE TRANSCRIPT]. Turn it into a [800]-word article for [AUDIENCE]. Pull out the three or four strongest ideas, quote the speaker directly where it lands well, cut the filler and tangents, and give it a clear intro and a takeaway ending. Keep the speaker's meaning intact and flag anything ambiguous. Output in Markdown.

Why it works: The 256K context window lets Kimi hold a long transcript at once, so it picks the best moments instead of just the first few.

4. Draft from a single idea

I want to write about this idea: [ONE-LINE IDEA]. Before drafting, ask me up to four questions to get the specifics you need (audience, angle, examples, length). Wait for my answers, then write the piece. Keep the tone [warm and specific], use short paragraphs, and avoid clichés and hype words.

Why it works: Starting from almost nothing, the clarifying questions force in the context Kimi needs before it writes.

Editing & rewriting

Paste your rough draft and Kimi shines at tightening, fixing flow, and cleaning up. Ask for the rewrite first and the change list second so the output stays clean but still shows its work.

5. Line edit for clarity

Line-edit the text below to be clearer and more direct without changing the meaning or my key terms. Cut filler, fix weak verbs, and break up any run-on sentences. Keep a [confident but plain] tone. Return the edited version first, then a short bullet list of the main changes you made and why. Text: [PASTE TEXT]

Why it works: Asking for the change list separately keeps the rewrite clean while still letting you see what moved.

6. Tighten a bloated draft

This draft is too long and loose. Cut it to [40%] shorter without losing any real point: [PASTE DRAFT]. Remove repetition, hedging, and throat-clearing, merge overlapping paragraphs, and keep every concrete fact and example. Return the tightened version, then tell me the two or three things you cut that I might miss.

Why it works: The flagged cuts let you rescue anything important when a draft says the right things in twice the words it needs.

7. Fix flow and transitions

The ideas in this piece are right but it reads choppy: [PASTE DRAFT]. Improve the flow between sentences and paragraphs, add transitions where they help, and reorder anything that's out of sequence. Don't add new content or change the facts. Return the smoother version, then note any reordering you did so I can check it.

Why it works: Scoping the task to flow only stops Kimi from rewriting content you were happy with.

8. Proofread and explain fixes

Proofread this text for grammar, spelling, punctuation, and consistency: [PASTE TEXT]. Fix the errors but leave my style and word choices alone. Return the corrected text, then a short list of the errors you found grouped by type. If something is a style choice rather than an error, leave it and say so.

Why it works: The "leave my style alone" instruction gives you a clean final pass that respects your voice instead of quietly rewriting it.

Tone & style

Voice comes from examples plus specific tone words. Give Kimi samples of your writing or your brand voice guide, or name the tone in plain language, and it will match closely. "Sound human" does nothing; "blunt, short sentences, no hype" does.

9. Match my writing voice

Here are three samples of my writing so you can learn my voice: [PASTE SAMPLES]. Now write [WHAT YOU NEED] in that same voice — match my sentence length, rhythm, vocabulary, and how direct I am. Keep the material I give you accurate: [PASTE MATERIAL]. Don't smooth it into generic corporate prose.

Why it works: Real samples give Kimi a concrete target to imitate, which beats any adjective you could pick to describe your style.

10. Shift the tone

Rewrite this text to shift the tone from [current tone] to [target tone], keeping the meaning and the facts exactly: [PASTE TEXT]. Match the new tone in word choice and sentence length, not just a few swapped words. Return the rewrite, then give me one alternate opening line in the same tone so I can compare.

Why it works: Naming both tones and asking for a real rewrite (not swapped words) dials a technically-correct piece to the right register.

11. Simplify to plain language

Rewrite this so a smart [general reader] gets it on the first read: [PASTE TEXT]. Replace jargon with plain words (or define it in one clause), shorten sentences, and use concrete nouns. Keep every fact and nuance — simpler, not dumbed down. Return the plain-language version and flag any term you had to keep because there's no simpler equivalent.

Why it works: "Simpler, not dumbed down" tells Kimi to preserve nuance while cutting jargon, which is the hard part of plain language.

12. Add wit without hype

This draft is accurate but flat: [PASTE DRAFT]. Give it a bit more personality and a lighter touch without changing the facts. Add dry wit where it fits naturally, vary the sentence rhythm, and sharpen the opening and closing lines. Do not add hype words, exclamation points, or forced jokes. Return the revised version.

Why it works: Banning hype and forced jokes warms up dry copy without tipping it into the over-eager AI voice.

Advertisement

Long-form (articles/reports)

Long-form is Kimi's strength — the 256K context holds a whole piece and its sources at once. Turn on Thinking and plan before you write. Build the outline first, then draft section by section so Kimi keeps the thread across the full length rather than losing it halfway.

13. Long-form outline first

Use Thinking mode. I'm writing a [2,000]-word piece on [TOPIC] for [AUDIENCE]. Before any prose, build a detailed outline: a working title, the core argument in one sentence, 5-7 H2 sections in a logical order, and 2-3 bullet points under each covering what it should say. Note where I'll need an example or data point. Don't write the article yet — just the outline.

Why it works: Thinking mode reasons through the structure while planning, so the outline reflects a real argument instead of a list of loosely related headings.

14. Section-by-section long draft

Here's the approved outline for my piece: [PASTE OUTLINE]. Write it one section at a time. Draft the first H2 section now — [300-400] words, practical tone, leading with the point and using the examples I noted. Stop at the end of the section and wait for me to say "next" before continuing, so I can steer as we go.

Why it works: Section-by-section keeps quality high across a long piece and stops the draft from drifting or thinning out toward the end.

15. Long report from source files

I've uploaded [several documents / a long brief]: [PASTE OR UPLOAD SOURCES]. Write a [1,500]-word report for [AUDIENCE] that synthesizes them. Structure: an executive summary up top, then themed sections that pull the key findings together, and a short list of open questions. Cite which source each key claim comes from, and don't add anything the sources don't support.

Why it works: The 256K window lets Kimi read every source file at once, so the report synthesizes across all of them instead of summarizing one at a time.

16. Build an argument with counterpoints

Use Thinking mode. Write a persuasive [800]-word piece arguing that [POSITION] for [AUDIENCE]. Build the case in a clear order, support each point with reasoning or the evidence I provide: [PASTE EVIDENCE]. Address the two strongest counterarguments honestly and answer them. Avoid strawmen and hype. End with a specific, concrete call to action.

Why it works: Thinking mode plans the logic, and forcing Kimi to face real objections makes the argument hold up instead of preaching to the choir.

Emails

For email, give Kimi the situation, the goal, and the tone, then keep it short. Instant mode is plenty here — you rarely need Thinking for a message under 200 words.

17. Professional email

Write a professional email from me ([MY ROLE]) to [RECIPIENT]. Situation: [CONTEXT]. Goal: [WHAT I WANT]. Keep it under [120] words, get to the point in the first line, stay [polite but direct], and end with one clear next step. Give me two subject-line options. Don't be stiff or over-formal.

Why it works: Naming the situation, goal, and word cap gives Kimi everything it needs to write a tight email in one pass.

18. Reply to a tricky message

I need to reply to this message: [PASTE MESSAGE]. My goal is [GOAL] and I want to come across as [calm and reasonable]. Draft a reply under [100] words that addresses their main point, holds my position where I need to, and keeps the relationship intact. If there's a sentence that could read as defensive, flag it and suggest a softer version.

Why it works: The defensive-line flag catches the phrasing you'd regret on a sensitive reply where the wording matters.

19. Cold outreach in your voice

Write a cold outreach email from me ([MY ROLE] at [MY COMPANY]) to [PROSPECT ROLE] at [PROSPECT COMPANY]. Goal: [GOAL]. Keep it under 90 words, open with a specific reason I'm reaching out to them (not a template line), state one clear value point, and end with a low-friction ask. Give me three subject lines. It should read like a person wrote it.

Why it works: The "specific reason, not a template" instruction is what separates outreach that gets a reply from mail-merge spam.

20. Summarize a thread into a reply

Here's a long email thread I've been added to: [PASTE THREAD]. First, summarize where things stand in three bullets: what's decided, what's open, and what people are waiting on. Then draft a reply from me that moves it forward — name the one decision we need and propose a next step. Keep the reply under [120] words.

Why it works: The 256K window handles even a long chain, so the summary is accurate and the reply responds to the whole thread in one go.

Social posts

Short-form writing rewards constraints. Give Kimi the platform, the audience, and a hard limit on length, and paste any real product detail or angle so the copy stays specific instead of vague.

21. Social post pack

Write [5] social posts about [TOPIC] for [PLATFORM], aimed at [AUDIENCE]. Vary the angle across the batch: a tip, a contrarian take, a short story, a question, and a stat. Keep each under [PLATFORM LIMIT] characters, no hashtag spam, no emoji spam, and make each one able to stand alone. Base them on this material: [PASTE MATERIAL].

Why it works: Asking for varied angles in one batch gives you a week of posts that don't all sound the same, grounded in your real material.

22. Headlines and hooks

Generate 12 headline options for [PIECE / PRODUCT] aimed at [AUDIENCE]. Mix angles: benefit-led, curiosity, contrarian, number-led, and how-to. Keep each under 65 characters, avoid hype words and clickbait, and label the angle of each. Then pick your top three and say in one line why they'd perform best for this audience.

Why it works: You get a batch of testable headlines with the reasoning attached, so you're choosing, not guessing.

Marketing copy

Marketing copy lives or dies on specifics. Paste the real product details, the buyer, and your brand voice, and ban the empty superlatives so every line earns its place.

23. Landing-page copy

Write landing-page copy for [PRODUCT] aimed at [BUYER]. Real details, features, and proof points: [PASTE DETAILS]. Match this brand voice: [PASTE BRAND VOICE OR SAMPLES]. Give me a hero headline and subhead, three benefit blocks (each a short heading plus two lines), one objection-handling section, and a primary call-to-action button label. No hype adjectives — every line should say something concrete.

Why it works: Pasting real proof points and your brand voice keeps the page specific and on-brand instead of generic SaaS filler.

24. Product description

Write a product description for [PRODUCT] aimed at [BUYER]. Here are the real details and features: [PASTE DETAILS]. Lead with the main benefit, translate each feature into what it does for the buyer, keep it to [LENGTH], and match a [tone] voice. No hype adjectives or empty superlatives — every line should say something concrete. Add a short bulleted spec list at the end.

Why it works: Pasting the real specs and banning hype forces Kimi to sell on concrete benefits instead of empty praise.

Want the shortest possible reference for all of this? The Kimi prompt cheat sheet puts the writing formula and every mode on one page, the Kimi prompt templates give you reusable fill-in-the-blank starters, and the best Kimi prompts roundup collects the full set across every use case.

Frequently Asked Questions

Which Kimi model are these writing prompts built for?

They target Kimi K2.7, the Moonshot AI flagship in the K2 family — a 1-trillion-parameter mixture-of-experts model (~32B active) with a 256K-token context window. That huge window is why you can paste a full brief, a long transcript, or an entire draft and still get a coherent result. It is free to start at kimi.com.

Does Kimi write well?

Yes, especially for long-form. Kimi K2.7 is a strong writer when given rich context: your notes, bullet points, brand voice, or an actual draft, plus a clear audience, tone, and length. The 256K context lets it hold long material at once, so it stays coherent across a full article. Its weakness is inventing facts from a thin prompt, so give it your material.

How do I get Kimi to write in my voice?

Paste two or three samples of your own writing, or your brand voice guide, and tell Kimi to match the rhythm, sentence length, and vocabulary. Then name the tone in plain words (blunt, warm, dry) and give it the real material to write about. Voice comes from examples plus specific tone words, not from asking it to "sound human".

Should I use Thinking or Instant mode for writing?

Leave Instant mode on for quick drafts, rewrites, emails, and short posts — it is faster and the quality is fine. Turn Thinking on when you need to structure a long or argument-heavy piece, because seeing the reasoning helps Kimi plan the outline before it writes.

Should I feed Kimi my notes or a finished draft?

Both work. Use file upload or paste to feed it notes, a brief, or a transcript when you want it to draft from scratch, and feed it your rough draft when you want editing, tightening, or a tone shift. Either way, giving it your real material keeps the facts yours and the output specific.

How do I keep Kimi from sounding generic?

Give it concrete material and ban the clichés. Add a line like "no filler, no hype words, lead each section with the point" and paste real examples, brand voice, or data. Generic output comes from thin prompts; specific input and a named tone fix most of it.

Can Kimi cite facts when it writes?

Yes. Turn on Web search and Kimi will pull current information and cite sources, which is useful when a piece needs a real stat or a fresh reference. For anything factual, ask it to cite the source inline so you can verify it before publishing.

How do the [BRACKET] placeholders work?

Anything in square brackets is a swap-in. Replace [AUDIENCE], [TOPIC], or [PASTE NOTES] with your own details before sending. The rest of each prompt is written to be copy-paste ready.

Advertisement