Kimi Prompts

I am uploading a very long document: [DOCUMENT TYPE, e.g. a 500 page industry report, a full novel manuscript] about [TOPIC]. Kimi's long context can handle it. First confirm the document length you received and give a 10 bullet map of its structure. Then answer my deep questions: [QUESTION 1], [QUESTION 2], [QUESTION 3], each with quoted evidence and section references. Finally: find 3 internal contradictions or inconsistencies in the document, and tell me what important angle on [TOPIC] the document completely ignores.

Act as my coding agent on this task: [TASK, e.g. add OAuth login to my Flask app, migrate these 3 modules to TypeScript]. Repo context: [DESCRIBE FILES OR PASTE KEY FILES]. Plan first: list every file you will touch and why, the order of operations, and how you will verify each step ([TEST COMMAND]). Then implement step by step across [NUMBER] tool calls without losing track. Rules: never rewrite unrelated code, keep [STYLE GUIDE], and stop and ask me before any destructive change like [EXAMPLE]. Summarize the final diff at the end.

I am uploading [NUMBER] long research papers on [TOPIC], each 30+ pages. Read all of them fully (use the long context, do not skim). Produce: a 600 word synthesis of the field's current state, a table comparing methods and findings with sample sizes, the 3 claims with the strongest evidence (quote them), the 2 most contested claims with both sides quoted, and the research gap my project on [MY PROJECT] should target. Cite papers as [Author Year] and give me a reading order from most to least important.

Research mission: I need a complete dossier on [COMPANY OR TOPIC] for [PURPOSE, e.g. a partnership decision, an investment memo]. Work agentically: search broadly, then drill into [ANGLE 1: financials], [ANGLE 2: reputation and reviews], [ANGLE 3: key people]. Deliver: a 2 page executive brief, a timeline of major events in the last [TIMEFRAME], red flags with sources, and 5 questions I should ask [STAKEHOLDER] in our meeting on [DATE]. Browse iteratively: if a lead looks promising, follow it before finalizing. List every source URL you used.

I am sharing a video about [TOPIC] ([LENGTH, e.g. a 2 hour lecture, a long podcast episode]). Process its full content and give me: the core thesis in 2 sentences, a detailed timestamped outline, the 5 strongest arguments with the speaker's own phrasing, anything the speaker gets wrong or oversimplifies about [SUBTOPIC] (with corrections), and whether it is worth my [TIME AVAILABLE] given my goal of [GOAL]. End with 3 resources that go deeper than this video.

Review this long [DOCUMENT TYPE, e.g. 80 page vendor agreement, terms of service] for [COMPANY OR CONTEXT]. I need: a plain language summary under 300 words, every clause about [CONCERN 1: e.g. data usage], [CONCERN 2: e.g. termination], [CONCERN 3: e.g. liability caps] extracted with exact quotes and section numbers, unusual clauses compared to standard [INDUSTRY] agreements, and a risk ranked list of what to negotiate. I am [YOUR ROLE]; highlight anything affecting [YOUR SPECIFIC INTEREST]. This is not legal advice; tell me when to hire a lawyer.

Plan the migration of [PROJECT DESCRIPTION] from [CURRENT STACK] to [TARGET STACK]. Codebase size: [LINES OR FILES], team: [TEAM SIZE], deadline: [DATE]. Deliver: phase by phase plan with effort estimates, which modules migrate first and why ([RISK ORDER]), compatibility shims needed for [DEPENDENCY], testing strategy per phase ([TEST TYPES]), rollback plan per phase, and the 5 highest risk items with mitigations. Ask me for any missing context about [UNKNOWN 1], [UNKNOWN 2] before finalizing the timeline.

Set reasoning effort to maximum. Problem: [HARD PROBLEM, e.g. optimize this logistics network, design a pricing model for my SaaS]. Context: [DATA, CONSTRAINTS, NUMBERS]. I need: your full reasoning trace summarized in clear steps (not just the answer), 3 candidate solutions compared on [CRITERION 1], [CRITERION 2], cost, and risk, sensitivity analysis on the key assumption ([ASSUMPTION]), and your recommended solution with implementation steps for the next [TIMEFRAME]. Challenge my constraints: which of [CONSTRAINT 1], [CONSTRAINT 2] should I relax?

Translate this long document from [SOURCE LANGUAGE] to [TARGET LANGUAGE]. Length: approximately [PAGES OR WORDS]. Requirements: preserve all formatting, headings, tables, and numbers exactly; keep these terms untranslated with explanations: [TERM 1], [TERM 2]; match the register ([FORMAL OR CASUAL]) of the original; flag idioms or culture specific references ([EXAMPLE]) with translator notes. After translating, give me a quality self check: 5 passages where translation was hardest and how you resolved each.

I am uploading [NUMBER] long reports ([TYPE, e.g. annual reports, audit documents]) from [ENTITIES]. Extract the same structured record from each: [FIELD 1], [FIELD 2], [FIELD 3], [FIELD 4]. Output as a markdown table, one row per report. Rules: quote the source passage for every extracted value, mark [NOT FOUND] explicitly instead of guessing, normalize [FIELD] to [FORMAT], and add a notes column for ambiguities. After the table, highlight the 3 most interesting cross report patterns you see in [METRIC].