Prompt Library · 2026

17+ AI Productivity Prompts

Ready-to-use prompt library for ChatGPT, Claude, and Gemini. Professional writing, meetings, data analysis, code, and brainstorming.

5 Techniques That Change Results

Mastering these techniques is worth more than memorizing 100 prompts.

1
Chain of Thought (CoT)when: Reasoning problems, math, logic

Add "Think step by step before answering" or "Show your reasoning"

"Calculate the campaign ROI. Think step by step and show each calculation."
2
Role Promptingwhen: Specialized perspective, consulting, technical analysis

Define the role before the task: "You are a [specialist]. Analyze..."

"You are an experienced CFO. Evaluate this investment proposal with healthy skepticism."
3
Few-Shot (examples)when: Tasks with a specific format, classification, extraction

Show 2–3 input→output examples before your real task

Input: "item arrived broken" → Category: Logistics | Input: "I want to cancel" → Category: Cancellation | Input: [your text]
4
Output Specificationwhen: Any task with an expected format

Specify exactly what you want: structure, length, tone, what NOT to include

"Respond in JSON with fields: title, summary (max 50 chars), tags (array). Do not include explanations outside the JSON."
5
Negative Promptingwhen: Avoiding frequently unwanted patterns

Explicitly tell the model what not to do

"Write the email. Do NOT use: 'Hope this finds you well', passive voice, or more than 3 paragraphs."
✍️

Professional Writing

Emails, reports, proposals, and corporate communications

Difficult email with assertiveness

When you need to give negative feedback or decline something

Write a professional, assertive email to [person/role] communicating that [difficult situation: proposal rejection / delay / negative feedback]. The tone should be direct but respectful, without excessive apologies. Include: (1) objective opening, (2) main message without hedging, (3) clear next steps. Maximum 150 words.
💡Replace the brackets with real context. Ask the model to "avoid phrases like Hope this finds you well".

Executive summary of a long document

Reports, contracts, or articles that need to be distilled for leadership

Read the text below and create an executive summary of no more than 200 words for a senior leadership audience without technical expertise. Structure it as: (1) Context in 1 sentence, (2) 3 main points in bullets, (3) Recommendation or next step. Prioritize business implications over technical details.

[Paste the document here]

Persuasive commercial proposal

Proposal for a client with known objections

Create a commercial proposal for [client/segment] offering [product/service] at [price]. The client has the following objections: [list the objections]. The proposal should: address each objection with data or argument, highlight 3 measurable tangible benefits, include legitimate urgency (not artificial), and end with a clear CTA. Tone: confident and consultative, not pushy.
💡The more real objections you list, the more relevant and persuasive the proposal becomes.

Transform a rough draft into polished prose

You have the ideas but the text is raw

Rewrite the text below preserving 100% of the information and my voice, but improving: clarity, paragraph structure, grammar, and flow. Do not add information that is not in the original. Do not make the text more formal than the original. If you find ambiguities, ask before inventing.

[Your draft here]
📅

Meetings & Planning

Agendas, meeting notes, follow-ups, and time management

Structured meeting minutes

After a meeting, turn raw notes into formal minutes

Turn the notes below into professional meeting minutes with: Date and attendees, Points discussed (concise bullets), Decisions made (bold), Next steps with owner and deadline (table), and Open items. Be objective and factual — do not invent missing information, mark it as [TO CONFIRM].

[Paste your notes here]

Efficient agenda to prevent unproductive meetings

An important meeting that needs a clear outcome

Create an agenda for a [duration] meeting on [topic/objective]. The meeting should end with [expected decision/deliverable]. Attendees: [list attendees and their roles]. Structure with: start time for each item, person responsible for leading, and what needs to be prepared in advance. Reserve the last 10% of the time for next steps.
💡Always include "what this meeting must produce" — it eliminates status meetings that could have been an email.

Follow-up that ensures action

After a meeting, ensure commitments are kept

Write a post-meeting follow-up email on [topic] with attendees [names]. The commitments made were: [list them]. The tone should be friendly but create accountability without being overbearing. Include: 3-line summary, action table (what/who/when), and an offer of support. The email subject line should be specific, not generic.
📊

Analysis & Data

Interpreting results, research, and synthesizing information

Data analysis without formulas

You have a table or CSV and need insights, not raw numbers

Analyze the data below and tell me: (1) The 3 main patterns or trends, (2) Anomalies or outliers that deserve attention, (3) A hypothesis to explain the most relevant pattern, (4) A question these data do not answer but would be important to investigate. Avoid repeating the raw numbers — focus on interpretation.

[Paste your table or data here]

Quick market research

You need a fast overview of a market or topic

Provide an analysis of [market/sector] with: estimated market size, main players and their positions, trends for the next 2 years, relevant barriers to entry, and an unexplored niche opportunity. Be specific and data-grounded. Cite limitations of what you know.

Argumentation with counterarguments

You need to defend a position or stress-test an idea

My position: [your position here]. First, present the 3 strongest arguments IN FAVOR of this position with evidence. Then, present the 3 strongest counterarguments from an intelligent critic. Finally, tell me which counterargument is hardest to rebut and how I could respond to it. Do not take sides — be an efficient devil's advocate.

Synthesis of multiple sources

You have read several articles or reports and need a coherent synthesis

Synthesize the texts below identifying: (1) What all sources agree on, (2) Where there is disagreement and why, (3) What is missing — perspectives not covered by any source, (4) Your most important conclusion considering the full set. If there are contradictions, point them out explicitly. Do not cite sources by number — integrate the ideas.

[Paste the texts here]
💻

Code & Technical

Debugging, documentation, review, and code generation

Debug with hypotheses

Code with unexpected behavior

My code has the following issue: [describe the bug — what you expected vs what happens]. Context: [language, framework, version]. Relevant code:

```
[paste the code]
```

Before suggesting a fix, list your 3 most likely hypotheses for the root cause, with a probability for each. Then fix based on the most likely hypothesis and explain why the solution works.
💡Asking for hypotheses before the fix reduces hallucinations and helps you understand the problem.

Production-focused code review

Before merging or deploying

Review the code below focusing on: (1) Bugs or edge cases that could cause production issues, (2) Performance problems at high volume, (3) Security issues (injection, data exposure, authentication), (4) Readability and maintainability. Do NOT suggest aesthetic refactors — only what impacts production, performance, or security.

```
[paste the code]
```

Technical documentation from code

Code with no documentation or outdated documentation

Generate technical documentation for the function/module below. Include: purpose in 1–2 sentences, parameters with type and description, return value, usage examples (happy path and edge cases), and known limitations. Write for a developer who does not know the code but knows the language.

```
[paste the code]
```
💡

Creativity & Brainstorming

Idea generation, naming, concepts, and problem solving

Brainstorming with real constraints

When "any idea" generates useless suggestions

Generate 10 ideas for [objective]. Mandatory constraints: (1) [constraint 1, e.g. budget < $5,000], (2) [constraint 2, e.g. team of 2 people], (3) [constraint 3, e.g. implementable in 30 days]. For each idea: name, 1-line description, why it respects the constraints. Do not filter by originality — include both obvious and non-obvious ideas. Then identify which has the most potential and why.
💡Real constraints are the secret to efficient brainstorming. Ideas without constraints rarely turn into action.

Name generation with criteria

A product, project, company, or feature that needs a name

Create 15 name options for [what it is: product/service/project] that [what it does] for [target audience]. Criteria: [e.g. memorable, max 2 syllables, .com domain available, no confusion with [competitor]]. Organize into 3 groups: descriptive names, evocative names, and invented names. Indicate which you would choose and why.

Problem solving with analogies

A complex problem that needs a fresh perspective

My problem: [describe the problem clearly]. Solve it using analogies from 3 different domains: (1) How would a software engineer solve it?, (2) How would a chef solve it?, (3) How would a military general solve it? For each analogy: what is the applicable principle and how does it translate to my problem. At the end, synthesize the most useful insight from all three.

How to Use This Library Efficiently

Generic prompts produce generic results. The secret is adapting templates with real context. For each prompt in this library, the most efficient cycle is:

  1. Use the prompt as-is to see the baseline response.
  2. Refine in the same conversation — ask the model to adjust tone, length, or angle without starting over. The conversation context is your ally.
  3. Save the adjusted prompt as a personal template for next time. Notion, Obsidian, or a simple text file all work.

For critical tasks (important communications, analyses that will be presented), always review the output as you would the work of a competent intern: are the facts correct? Is the tone right? Did the model invent or over-generalize anything? AI is an accelerator, not a substitute for your judgment.

Frequently Asked Questions

What is the difference between prompts for ChatGPT, Claude, and Gemini?
All three models respond well to the same well-written prompts, but each has nuances. Claude follows instructions very literally and excels at analyzing long texts and tasks with many constraints. GPT-4o is more conversational and handles roleplay and creativity well. Gemini has advantages on tasks requiring live data (via Google Search) and multimodal inputs. For most professional prompts, the difference is small — a well-written prompt works across all three.
Why do my prompts generate generic responses?
Generic responses are a symptom of insufficient context. The 4 elements that most improve specificity: (1) who you are and your role, (2) who the output is for and what they already know, (3) real constraints (length, tone, what not to include), (4) expected format with an example. A 3-line prompt with these elements produces far better responses than long but vague prompts.
Should I use the same prompt or adjust it for each conversation?
Create templates for recurring tasks. A well-calibrated "meeting minutes" prompt is worth saving and reusing — just adjust the variables (attendees, topic, date). For unique and complex tasks, start with a robust prompt and iterate in the same conversation: ask for alternative versions, tone adjustments, format variations. The conversation history is free context.
Is prompt engineering a difficult skill to learn?
The learning curve is fast. Within 2–3 hours of practice with the fundamentals (clear context, explicit constraints, expected format, examples), most people can produce prompts that generate useful results. What takes practice is calibrating the level of detail: too little context → generic response; irrelevant details → the model gets distracted. The highest-impact technique for beginners is output specification: describing exactly what you want to see in the response.
How do I write effective prompts in a language other than English?
Modern LLMs (GPT-4o, Claude, Gemini) are highly proficient in most major languages. For best results: (1) write your prompt in your target language — the model will respond in the same language, (2) specify "respond in British English" or the exact regional variant if needed, (3) for technical jargon, you can mix languages without issue — models handle code-switching fluently. Native-language prompts consistently outperform auto-translated English ones.