These are 18 fill-in-the-blank templates for Kimi, the Moonshot AI assistant that is free to start at kimi.com. Each one is a reusable skeleton: replace the parts in [BRACKETS] with your own details and paste the whole thing in. They target the Kimi K2.7 / K2 family — a 1-trillion-parameter mixture-of-experts model (about 32B active) with a 256K context window and multimodal input — and they are grouped by job: coding, Agent Mode, writing, research, business, and study.

Every template is built on the same shape: Role + Task + Context + Constraints + Output format. A few habits make them land harder. Switch from Instant to Thinking for coding, research, and multi-step work so you can watch the reasoning; turn on Web search (or Deep Research) for the research ones and demand citations; and hand the full brief to Agent Mode / OK Computer when you want Kimi to actually build the deliverable. New here? The best Kimi prompts roundup is the full set, and the Kimi prompt cheat sheet is a one-page reference.

Advertisement

Coding

Switch to Thinking for hard bugs and design work. Always paste the real code, name your language and versions, and give the exact error. Kimi's 256K context means you can paste a whole file, not a snippet.

1. Debug with a root-cause trace

You are a senior [LANGUAGE] engineer. Turn on Thinking. I'm getting this error in [LANGUAGE / FRAMEWORK + version]: [PASTE ERROR].
Here's the relevant code: [PASTE CODE].
What I expected: [EXPECTED]. What happens instead: [ACTUAL]. What I've already tried: [ATTEMPTS].
Trace the root cause step by step. If more than one cause is possible, rank them by likelihood. Give me the corrected code and explain the fix in two sentences, then note any edge case it might miss.

Fill in: [LANGUAGE / FRAMEWORK + version], [PASTE ERROR], [PASTE CODE], [EXPECTED], [ACTUAL], [ATTEMPTS].

2. Build a function from a spec

You are a [LANGUAGE] developer. Write a function that does this: [DESCRIBE BEHAVIOR].
Inputs: [INPUTS AND TYPES]. Outputs: [OUTPUT]. Edge cases to handle: [EDGE CASES]. Constraints: [PERFORMANCE / DEPENDENCIES / STYLE].
Write clean, idiomatic code with brief comments on the tricky parts. Then walk through each edge case above to confirm it holds, and list any input you're unsure about.

Fill in: [LANGUAGE], [DESCRIBE BEHAVIOR], [INPUTS AND TYPES], [OUTPUT], [EDGE CASES], [PERFORMANCE / DEPENDENCIES / STYLE].

3. Refactor without changing behavior

Refactor this code for readability and maintainability without changing behavior: [PASTE CODE].
Improve naming, reduce nesting, and extract logic where it helps. Keep it idiomatic for [LANGUAGE] and don't add dependencies beyond [ALLOWED DEPENDENCIES].
Return the refactored version, then a short bullet list of what you changed and why. As a final check, confirm the inputs and outputs still match the original for the same cases.

Fill in: [PASTE CODE], [LANGUAGE], [ALLOWED DEPENDENCIES].

Agent Mode (OK Computer)

These hand the full brief to Agent Mode (branded OK Computer), where Kimi's Agent Swarm plans, executes, and assembles a finished deliverable — a website, deck, doc, or sheet. Be specific about the output and the constraints; the more you pin down, the less it guesses. For more, see the Kimi agent prompts.

4. Build a website with Agent Mode

Use Agent Mode (OK Computer). Build a [TYPE OF SITE, e.g. landing page / portfolio / docs site] for [PROJECT / BUSINESS].
Audience: [AUDIENCE]. Goal of the site: [GOAL / DESIRED ACTION].
Pages / sections: [LIST SECTIONS]. Content to use: [PASTE COPY OR KEY POINTS]. Style: [BRAND / COLORS / TONE / REFERENCE SITES].
Constraints: [RESPONSIVE / SINGLE PAGE / NO BACKEND / FRAMEWORK]. Deliver the working site, then list what you built and how to edit each section.

Fill in: [TYPE OF SITE], [PROJECT / BUSINESS], [AUDIENCE], [GOAL], [LIST SECTIONS], [PASTE COPY], [STYLE], [CONSTRAINTS].

5. Design-to-code from a mock

Use Agent Mode (OK Computer). I've uploaded a design mock: [UPLOAD SCREENSHOT / FIGMA EXPORT / IMAGE].
Turn it into a working [FRAMEWORK, e.g. HTML/CSS / React] front end that matches the layout, spacing, and colors as closely as possible.
Make it responsive down to [BREAKPOINT]. Use [FONT / COLOR TOKENS] where I've specified them, otherwise infer from the mock. Keep the markup clean and semantic.
Deliver the code, then note any spot where the mock was ambiguous and what you assumed.

Fill in: [UPLOAD SCREENSHOT / IMAGE], [FRAMEWORK], [BREAKPOINT], [FONT / COLOR TOKENS].

6. Build a slide deck

Use Agent Mode (OK Computer). Build a [NUMBER]-slide deck on [TOPIC] for [AUDIENCE].
Purpose: [PERSUADE / INFORM / PITCH]. Key points to cover, in order: [POINT 1], [POINT 2], [POINT 3]. Source material: [PASTE OR UPLOAD].
Style: [CLEAN / BOLD / CORPORATE + any brand colors]. One idea per slide, a clear headline on each, minimal text, and speaker notes underneath. Deliver the deck and an outline of the slide flow.

Fill in: [NUMBER], [TOPIC], [AUDIENCE], [PURPOSE], [POINT 1/2/3], [SOURCE MATERIAL], [STYLE].

Writing

Kimi matches your voice well when you give it a Role, real Context, and an exact Output format. Paste your actual notes so the facts stay yours and it only handles structure and polish.

7. Draft from an outline

You are a writer for [AUDIENCE]. Write a [LENGTH]-word [PIECE TYPE] titled "[TITLE]" from this outline: [PASTE OUTLINE].
Tone: [TONE]. Lead each section with the takeaway, use short paragraphs and concrete examples, and cut clichés and filler. Output in clean Markdown with H2 sections and a two-line intro.

Fill in: [AUDIENCE], [LENGTH], [PIECE TYPE], [TITLE], [PASTE OUTLINE], [TONE].

8. Rewrite for clarity and tone

Rewrite the text below to be clearer and more direct without losing meaning. Cut filler, fix weak verbs, and keep my key terms.
Target tone: [TONE]. Reading level: [LEVEL].
Return the rewrite first, then a short bullet list of the main changes you made.
Text: [PASTE TEXT]

Fill in: [TONE], [LEVEL], [PASTE TEXT].

9. Repurpose one piece into many

Repurpose this piece into content for other channels: [PASTE ARTICLE / TRANSCRIPT].
I want: [NUMBER] short social posts, a [LENGTH]-word newsletter blurb, and [NUMBER] headline options.
Keep my voice [TONE], stay faithful to the original points, and don't invent facts that aren't in the source. Label each output clearly.

Fill in: [PASTE ARTICLE / TRANSCRIPT], [NUMBER], [LENGTH], [TONE].

Advertisement

Research & analysis

Turn on Web search for this section so Kimi pulls live pages instead of answering from memory, or use Deep Research for a longer cited report. Always ask it to cite every source with a link so you can verify each claim.

10. Cited research brief

Turn on Web search. Research this question and brief me: [RESEARCH QUESTION].
Cover: the current state, the main viewpoints, and where the evidence points. Prefer primary and recent sources.
Cite every source with a link, note the publication date next to each key fact, and end with a short synthesis of where sources agree and where they conflict.

Fill in: [RESEARCH QUESTION].

11. Deep Research report

Use Deep Research. Produce a report on [TOPIC / QUESTION] for a [WHO] reader who needs to [DECISION / GOAL].
Break it into: background, the key findings, the strongest counter-evidence, and a bottom-line recommendation. Prioritize recent, primary, and credible sources.
Cite every claim with a linked source and its date. At the end, list the open questions and what would resolve them.

Fill in: [TOPIC / QUESTION], [WHO], [DECISION / GOAL].

12. Compare options with sources

Turn on Web search. Compare [OPTION A] and [OPTION B] for a [USE CASE] buyer.
Build a table: price, key features, real user complaints, and best-fit user. Pull from current reviews and forums, and cite every source with a link.
End with a clear recommendation for someone who cares most about [PRIORITY], and flag anything the sources disagreed on.

Fill in: [OPTION A], [OPTION B], [USE CASE], [PRIORITY].

Business & work

Give Kimi your real context and it handles the structure. Switch to Thinking when the task needs judgment, like pressure-testing a plan or triaging priorities.

13. One-page plan for an idea

You are a pragmatic advisor. Draft a one-page plan for this idea: [DESCRIBE IDEA].
Cover: the problem, the target customer [TARGET CUSTOMER], the offer, how it makes money, the first three steps to validate it cheaply, and the biggest risk. Be blunt about weaknesses. Output as clear sections with no filler.

Fill in: [DESCRIBE IDEA], [TARGET CUSTOMER].

14. Meeting notes to action items

Turn these meeting notes into a clean summary: [PASTE OR UPLOAD NOTES].
Output three sections: Decisions made, Action items (as a table with Owner, Task, Due date), and Open questions. Keep it tight, don't invent details that aren't in the notes, and flag anything left unresolved.

Fill in: [PASTE OR UPLOAD NOTES].

15. Professional email drafter

Write an email from me ([MY ROLE]) to [RECIPIENT] about [TOPIC].
Goal: [WHAT I WANT TO HAPPEN]. Tone: [TONE]. Keep it under [WORD COUNT] words.
Open with the reason I'm writing, state the ask clearly, and end with a specific next step. Give me two subject-line options.

Fill in: [MY ROLE], [RECIPIENT], [TOPIC], [WHAT I WANT TO HAPPEN], [TONE], [WORD COUNT].

Study & learning

Kimi is a strong tutor when you tell it your level and what to do with a wrong answer. Switch to Thinking for step-by-step explanations of hard concepts, and upload your notes so it works from your material.

16. Explain a hard concept

Explain [CONCEPT] to me. I currently know [YOUR LEVEL / WHAT YOU ALREADY UNDERSTAND].
Start with a plain-language explanation, then one concrete everyday analogy, then a worked example step by step. Finish with the single most common misconception about it and why it's wrong. Keep it concise and don't assume knowledge I didn't mention.

Fill in: [CONCEPT], [YOUR LEVEL / WHAT YOU ALREADY UNDERSTAND].

17. Study guide and quiz maker

Make a study guide for [TOPIC / EXAM] at a [LEVEL] level from this material: [PASTE OR UPLOAD NOTES].
Give me: the key concepts as a bulleted outline, 8 flashcard-style Q&A pairs, and a 6-question quiz (mixed difficulty) with an answer key at the bottom. Base everything on the material provided, and mark anything you added from general knowledge.

Fill in: [TOPIC / EXAM], [LEVEL], [PASTE OR UPLOAD NOTES].

18. Socratic tutor

Be my Socratic tutor for [SUBJECT / SKILL]. My level: [LEVEL]. My goal: [GOAL].
Don't just give answers — ask me one guiding question at a time, wait for my reply, and adjust based on what I get right or wrong. When I'm stuck, give a hint before the answer. Keep each turn short and end each one with the next question.

Fill in: [SUBJECT / SKILL], [LEVEL], [GOAL].

Want the shortest reference for all of this? The Kimi prompt cheat sheet lists every mode and modifier on one page, and the best Kimi prompts roundup has the full set worth saving.

Frequently Asked Questions

How do the fill-in-the-blank templates and brackets work?

Each template is a skeleton built on Role + Task + Context + Constraints + Output format. Anything in square [BRACKETS] is a swap-in: replace [TOPIC], [PASTE CODE], or [AUDIENCE] with your own details, then paste the whole thing into Kimi. The wording around the brackets is written to stay copy-paste ready.

Which Kimi model do these work with?

They target the Kimi K2.7 / K2 family, the free flagship at kimi.com — a 1-trillion-parameter mixture-of-experts model (about 32B active) with a 256K context window and multimodal input. Nothing is version-specific, so they also work on earlier K2 releases and via the API.

When should I turn on Thinking mode?

Switch from Instant to Thinking for the coding, research, and multi-step business templates, and any time the problem has several constraints. Thinking shows the reasoning so you can catch a wrong turn. For quick rewrites or lookups, Instant is faster.

What is Agent Mode / OK Computer?

Agent Mode (branded OK Computer) lets Kimi autonomously build deliverables — websites, docs, slide decks, and spreadsheets — using an Agent Swarm that plans, executes, and assembles the result. The agent templates above hand it a full brief so it can run end to end.

When should I turn on Web search or Deep Research?

Turn on Web search for the research and analysis templates, or anytime you need current facts. Deep Research goes further and returns a cited report. Add "cite every source with a link" so the output stays auditable.

Can I combine templates?

Yes. Run a research template with Web search to gather cited facts, then feed the result into a writing template, or hand the whole brief to Agent Mode to build the deliverable. Kimi's 256K context holds plenty, so you can paste long inputs and chain steps in one chat.

Are these templates free to use?

Yes. Kimi is free to start at kimi.com, and these templates are free to copy, edit, and reuse. No paid tier is required for any of them.

Can Kimi read files and images?

Yes. Kimi is multimodal and supports file and image upload, so several templates say "paste or upload". You can hand it a PDF, a spreadsheet, a screenshot, or a design mock and it will work from the source — design-to-code is a supported use case.

Advertisement