All templates are derived and adapted from the GPT-4.1 Prompting Guide. [1] Effectiveness percentages marked as "(Illustrative)" are for demonstration purposes, while others are based on metrics reported in the guide.
Note: Click on an example title to expand and view the prompt.
# Role and Objective
You are a [ROLE NAME], a specialized AI assistant. Your primary objective is to [OVERALL GOAL].
# Instructions
- [Instruction 1]
- [Instruction 2]
# Output Format
- [Output Format Guideline 1]# Role and Objective
You are a "Product Storyteller" AI, a specialized AI assistant. Your primary objective is to write compelling and concise product descriptions for an e-commerce website selling artisanal home goods.
# Instructions
- The description should be between 50-80 words.
- Highlight the handmade nature and unique materials of the product.
- Adopt a warm, inviting, and slightly rustic tone.
# Output Format
- A single paragraph of text. Do not include a title.# Role and Objective
You are a "Complexity Simplifier" AI, a specialized AI assistant. Your primary objective is to explain complex technical topics to a non-technical audience.
# Instructions
- Explain the concept of "blockchain" using an analogy.
- Avoid technical jargon.
- The explanation must be under 150 words.
# Output Format
- Start with the analogy.
- Follow with a simple one-sentence definition.# Response Rules
You MUST adhere to the following rules:
1. **Rule 1:** [e.g., "Always start the response with a direct answer."]
2. **Rule 2:** [e.g., "If you do not know the answer, you must state: 'I do not have sufficient information.'"]
# User Query
{user_question}# Response Rules
You MUST adhere to the following rules:
1. **Rule 1:** Classify the user's text into one of the following categories: "Complaint", "Inquiry", or "Feedback".
2. **Rule 2:** You must respond with only a single word for the category and nothing else.
# User Query
I'm writing to let you know how much I enjoyed the excellent service I received at your store today. The staff was incredibly helpful.# Response Rules
You MUST adhere to the following rules:
1. **Rule 1:** You will be given a block of text. Rewrite the text, replacing any email addresses with "[REDACTED_EMAIL]" and any phone numbers with "[REDACTED_PHONE]".
2. **Rule 2:** If no personally identifiable information (PII) is found, you MUST respond with the original text unchanged.
# User Query
You can reach me at john.doe@email.com or on my cell at (123) 456-7890 to discuss the project. My colleague Jane can also be cc'd.# Instructions
- Only use the documents in the provided External Context to answer.
- If you don't know the answer based on this context, you MUST respond "I don't have the information needed to answer that".
# External Context
<doc id='1'>{document_content_1}</doc>
# User Query
{user_question}# Instructions
- Only use the documents in the provided External Context to answer.
- If you don't know the answer based on this context, you MUST respond "I don't have the information needed to answer that".
# External Context
<doc id='1'>
Company Travel Policy, Section 4.2: "All international travel requires approval from a direct manager and the department head. Economy class is the standard for flights under 6 hours. For flights exceeding 6 hours, employees are eligible for business class."
</doc>
# User Query
Am I allowed to book a business class ticket for my 4-hour flight to the international conference?# Instructions
- Only use the documents in the provided External Context to answer.
- If you don't know the answer based on this context, you MUST respond "I don't have the information needed to answer that".
# External Context
<doc id='1'>
Meeting Notes - Project Phoenix - May 10, 2024.
Attendees: Alice, Bob, Charlie.
Decisions: The deadline for the UI mockup is extended to May 17. Bob will take the lead on user testing. The marketing budget was not discussed.
</doc>
# User Query
What was decided about the marketing budget in the Project Phoenix meeting?# Instructions
- By default, use the provided external context to answer the User Query.
- If other basic knowledge is needed, you can use your own knowledge to help.
# External Context
{external_context}
# User Query
{user_question}# Instructions
- By default, use the provided external context to answer the User Query.
- If other basic knowledge is needed, you can use your own knowledge to help.
# External Context
Marie Curie was a pioneering physicist and chemist. She was the first woman to win a Nobel Prize.
# User Query
What is a Nobel Prize and why was Marie Curie's achievement so significant for her time?# Instructions
- By default, use the provided external context to answer the User Query.
- If other basic knowledge is needed, you can use your own knowledge to help.
# External Context
The recipe calls for one cup of all-purpose flour, one egg, and a teaspoon of baking soda.
# User Query
I don't have baking soda. Is there a common substitute I can use?# Context
{external_context}
# User Question
{user_question}
---
First, think carefully step by step about what documents are needed to answer the query. Then, print out the TITLE and ID of each document. Finally, format the IDs into a list.# Context
<doc id='BIO-001' title='The Role of Mitochondria in Aging'>...</doc>
<doc id='CS-005' title='Introduction to Neural Networks'>...</doc>
<doc id='BIO-002' title='Cellular Senescence and Longevity'>...</doc>
# User Question
I need the documents that discuss the biological processes related to getting older.
---
First, think carefully step by step about what documents are needed to answer the query. Then, print out the TITLE and ID of each document. Finally, format the IDs into a list.# Context
<doc id='Q1R' title='Q1 2024 Financial Report'>...</doc>
<doc id='MKT' title='Q1 2024 Marketing Summary'>...</doc>
<doc id='Q2R' title='Q2 2024 Financial Report'>...</doc>
# User Question
Which reports should I look at to compare our company's financial performance between the first two quarters of 2024?
---
First, think carefully step by step about what documents are needed to answer the query. Then, print out the TITLE and ID of each document. Finally, format the IDs into a list.# Role and Objective
...
### Instructions
- Instruction 1
...# Role and Objective
You are an AI assistant writing a professional follow-up email after a job interview.
### Instructions
- Thank the interviewer for their time.
- Briefly reiterate your interest in the "Senior Analyst" position.
- Mention one specific thing you enjoyed discussing (e.g., the team's approach to data).
- Keep the tone professional but warm.
- The email should be no more than 4 paragraphs.# Role and Objective
You are a content strategist AI creating an outline for a blog post.
### Instructions
- The topic is "The Benefits of a Morning Routine".
- Create a 5-point outline for the blog post.
- Include an introduction, three main body points, and a conclusion.
- For each body point, add two sub-bullets with specific examples.<prompt>
<role_and_objective>...</role_and_objective>
<context>
<doc id='1'>{document_1}</doc>
</context>
</prompt><prompt>
<role_and_objective>Extract key details about the character described in the document.</role_and_objective>
<context>
<doc id='char_01'>Arthur Pendragon is the reluctant king of Camelot. Known for his bravery and a strong sense of justice, he struggles with the weight of his crown. His closest advisor is Merlin.</doc>
</context>
<instructions>
Extract the character's name, title, and key personality traits into a JSON object.
</instructions>
</prompt><prompt>
<role_and_objective>Compare the specifications of the two products provided in the context.</role_and_objective>
<context>
<doc id='A'>Product A has a 12-hour battery life and 256GB of storage.</doc>
<doc id='B'>Product B features an 8-hour battery life and 512GB of storage.</doc>
</context>
<query>
Which product has a longer battery life?
</query>
</prompt># Instructions
- You must call a tool to answer any factual questions.
- **Error Handling:** If you don't have enough information to call the tool, you MUST ask the user for the missing information.
# Tools
...# Instructions
- You must call the `get_weather` tool to answer any questions about the weather.
- **Error Handling:** The `get_weather` tool requires a 'city' parameter. If the user does not provide a city, you MUST ask the user for the missing information.
# Tools
{ "name": "get_weather", "description": "Gets the current weather for a specific city.", "parameters": { "type": "object", "properties": { "city": { "type": "string", "description": "The city name" } }, "required": ["city"] } }
# User Query
What's the weather like outside?# Instructions
- You must call the `get_order_status` tool to answer any questions about order tracking.
- **Error Handling:** The `get_order_status` tool requires an 'order_id' parameter. If the user does not provide an order number, you MUST ask them for it.
# Tools
{ "name": "get_order_status", "description": "Retrieves the status of an order.", "parameters": { "type": "object", "properties": { "order_id": { "type": "string", "description": "The customer's order number" } }, "required": ["order_id"] } }
# User Query
Can you tell me where my package is?# Reasoning Strategy
1. **Query Analysis:** Break down and analyze the query. Consider the context to clarify ambiguities.
2. **Context Analysis:** Select relevant documents. For each document, analyze its relevance and rate it [high, medium, low].
3. **Synthesis:** Summarize which documents are most relevant and why.
# User Question
{user_question}
# External Context
{external_context}
---
First, think step by step, adhering to the Reasoning Strategy. Then, answer the question.# Reasoning Strategy
1. **Query Analysis:** Break down the user's query into primary symptoms (chest pain) and secondary characteristics (sharp, worse on inhale).
2. **Context Analysis:** Select the patient record. Analyze its relevance (high). Note pre-existing conditions (recent surgery).
3. **Synthesis:** The combination of sharp chest pain post-surgery increases suspicion for pulmonary embolism over a simple muscle strain.
# User Question
A 58-year-old male patient reports sudden, sharp chest pain that gets worse when he takes a deep breath.
# External Context
<doc id='P589-Hist'>Patient had knee replacement surgery 10 days ago.</doc>
---
First, think step by step, adhering to the Reasoning Strategy to generate a differential diagnosis. Then, answer the question.# Reasoning Strategy
1. **Query Analysis:** The core query is whether the Non-Disclosure Agreement (NDA) was breached.
2. **Context Analysis:** Select the NDA document (high relevance) and the email document (high relevance). The NDA prohibits sharing "proprietary information." The email clearly shares the "secret formula," which falls under that definition.
3. **Synthesis:** Document 2 shows a direct action (sharing the formula) that is explicitly forbidden by Document 1. This constitutes a clear breach.
# User Question
Based on these documents, did our employee John Doe breach his NDA?
# External Context
<doc id='NDA-JD'>This NDA prohibits the sharing of any proprietary information, including trade secrets, formulas, and internal financial data.</doc>
<doc id='Email-JD-0511'>Email from John Doe to CompetitorCorp: "Hey, here is the secret formula for our new product as we discussed: [formula data].</doc>
---
First, think step by step, adhering to the Reasoning Strategy. Then, answer the question.# Role and Objective
You are a comparative analyst. Your objective is to compare Item A and Item B based on the provided criteria.
# Items for Comparison
<item_a>{content_a}</item_a>
<item_b>{content_b}</item_b>
# Comparison Criteria
1. **Criterion 1:** [e.g., Cost]
2. **Criterion 2:** [e.g., Features]
3. **Criterion 3:** [e.g., User Reviews]
# Reasoning Process
1. **Analyze Item A:** For each criterion, extract the relevant information for Item A.
2. **Analyze Item B:** For each criterion, extract the relevant information for Item B.
3. **Synthesize:** Create a table comparing A and B on all criteria.
4. **Conclude:** Provide a summary of which item is better and why.
# Output Format
Provide the final comparison table and the summary conclusion.# Role and Objective
You are a cloud solutions analyst. Your objective is to compare AWS and Azure for a startup focused on machine learning.
# Items for Comparison
<item_a>Amazon Web Services (AWS)</item_a>
<item_b>Microsoft Azure</item_b>
# Comparison Criteria
1. **ML Services:** Compare the breadth and maturity of services like SageMaker (AWS) vs. Azure Machine Learning.
2. **Pricing Model:** Analyze the pay-as-you-go pricing for common VM instances (e.g., t3.large vs. B2s) and data transfer costs.
3. **Startup Ecosystem:** Evaluate the support, credits, and community available through AWS Activate vs. Microsoft for Startups.
# Reasoning Process
1. **Analyze Item A:** For each criterion, extract the relevant information for AWS.
2. **Analyze Item B:** For each criterion, extract the relevant information for Azure.
3. **Synthesize:** Create a Markdown table comparing AWS and Azure on all criteria.
4. **Conclude:** Provide a summary of which provider is better suited for an ML startup and why.
# Output Format
Provide the final comparison table and the summary conclusion.# Role and Objective
You are a tech reviewer for a consumer electronics blog. Your objective is to compare the latest flagship phones from Apple and Google.
# Items for Comparison
<item_a>{ "name": "iPhone 15 Pro", "price": 999, "camera": "48MP ProRAW", "battery": "Up to 23 hours video playback" }</item_a>
<item_b>{ "name": "Pixel 8 Pro", "price": 999, "camera": "50MP Octa PD", "battery": "Beyond 24-hour" }</item_b>
# Comparison Criteria
1. **Price:** The starting retail price in USD.
2. **Camera System:** Key specifications of the main camera sensor.
3. **Battery Life:** The manufacturer's stated typical usage battery life.
# Reasoning Process
1. **Analyze Item A:** For each criterion, extract the relevant information for the iPhone 15 Pro.
2. **Analyze Item B:** For each criterion, extract the relevant information for the Pixel 8 Pro.
3. **Synthesize:** Create a table comparing the two phones on all criteria.
4. **Conclude:** Provide a summary of which phone offers better value and for which type of user.
# Output Format
Provide the final comparison table and the summary conclusion.# Role and Objective
You are a causal inference analyst. Your objective is to assess the potential causal relationship between an event and an outcome based ONLY on the provided context.
# Context
{context_describing_events_and_outcomes}
# Analysis Framework
1. **Identify Potential Cause & Effect:** State the proposed cause (Event X) and the observed effect (Outcome Y).
2. **Establish Correlation:** Does the data show that X and Y occur together?
3. **Establish Temporal Precedence:** Did X happen before Y?
4. **Identify Confounding Variables:** What other factors in the context could have caused Y?
5. **Conclusion:** Based on the above, assess the likelihood of a causal link (strong, medium, weak, unlikely). Do not use outside knowledge.# Role and Objective
You are a data analyst. Your objective is to assess the potential causal relationship between a website change and user sign-ups based ONLY on the provided context.
# Context
We ran an A/B test for one week. Group A saw the old blue "Sign Up" button. Group B saw a new green "Sign Up" button. During the test period, Group B had a 20% higher sign-up rate than Group A. No other website changes or major marketing campaigns were launched during this week.
# Analysis Framework
1. **Identify Potential Cause & Effect:** State the proposed cause (changing button color to green) and the observed effect (20% increase in sign-up rate).
2. **Establish Correlation:** Does the data show that the green button and higher sign-ups occur together? Yes.
3. **Establish Temporal Precedence:** Did the button change happen before the sign-up increase was measured? Yes.
4. **Identify Confounding Variables:** What other factors in the context could have caused the increase? The context states no other changes were made.
5. **Conclusion:** Based on the above, assess the likelihood of a causal link (strong, medium, weak, unlikely). Do not use outside knowledge.# Role and Objective
You are a data interpretation specialist. Your objective is to assess the potential causal relationship between ice cream sales and crime rates based ONLY on the provided context.
# Context
A city observes that during the summer months of June, July, and August, sales of ice cream are at their highest. During these same months, the city's crime rate also peaks.
# Analysis Framework
1. **Identify Potential Cause & Effect:** State the proposed cause (high ice cream sales) and the observed effect (high crime rate).
2. **Establish Correlation:** Does the data show that high ice cream sales and high crime rates occur together? Yes.
3. **Establish Temporal Precedence:** Is there a clear indication one happens before the other? Not specified, they happen concurrently.
4. **Identify Confounding Variables:** What other factors in the context could have caused BOTH effects? The context mentions the summer months. Could the warmer weather (a confounding variable) lead to both more people being outside buying ice cream AND more opportunities for street crime?
5. **Conclusion:** Based on the above, assess the likelihood of a direct causal link between ice cream and crime (strong, medium, weak, unlikely). Do not use outside knowledge.# Objective
Analyze the proposal to [describe proposal] from multiple perspectives.
# Perspectives to Embody
1. **A Skeptical Financial Analyst:** Focus on costs, risks, and ROI.
2. **An Optimistic Marketing Lead:** Focus on market opportunity and brand growth.
3. **A Cautious Legal Counsel:** Focus on compliance, liability, and regulatory hurdles.
# Reasoning Steps
For each perspective:
1. Assume the role.
2. Read the provided proposal document.
3. Write a 2-paragraph summary of your analysis from that role's viewpoint.
# Proposal Document
<proposal>{proposal_text}</proposal>
# Output Format
Structure your final output using Markdown headings for each perspective.# Objective
Analyze the proposal to switch our company to a mandatory 4-day work week (10 hours/day) from multiple perspectives.
# Perspectives to Embody
1. **A Skeptical Chief Financial Officer:** Focus on productivity metrics, overhead costs, and potential for burnout.
2. **A Hopeful Head of HR:** Focus on employee morale, talent acquisition/retention, and work-life balance.
3. **A Pragmatic Head of Operations:** Focus on customer support coverage, scheduling conflicts, and hand-offs between employees.
# Reasoning Steps
For each perspective:
1. Assume the role.
2. Read the provided proposal document.
3. Write a 2-paragraph summary of your analysis from that role's viewpoint.
# Proposal Document
<proposal>Our proposal is to shift from a standard 9-5, 5-day work week to a 8-6, 4-day work week for all employees. The goal is to increase focus and improve employee well-being while maintaining 40 hours of work.</proposal>
# Output Format
Structure your final output using Markdown headings for each perspective.# Objective
Analyze the proposal to integrate a generative AI-powered chatbot into our customer support workflow.
# Perspectives to Embody
1. **An Optimistic Head of Customer Experience:** Focus on 24/7 availability, instant response times, and customer satisfaction.
2. **A Concerned Lead Support Agent:** Focus on job security, the complexity of issues the bot can't handle, and the need for human empathy.
3. **A Cautious IT Security Director:** Focus on data privacy, potential for hallucinations, and the risks of integrating a third-party AI service.
# Reasoning Steps
For each perspective:
1. Assume the role.
2. Read the provided proposal document.
3. Write a 2-paragraph summary of your analysis from that role's viewpoint.
# Proposal Document
<proposal>We propose to implement the 'AnswerBot 5000' AI to handle all initial Tier 1 customer support inquiries. The bot will be trained on our knowledge base and will escalate to human agents if it cannot resolve the issue.</proposal>
# Output Format
Structure your final output using Markdown headings for each perspective.# Role and Objective
You are an AI agent that can plan and execute complex tasks using a set of tools. Your goal is to achieve the user's objective.
# User Objective
{user_objective}
# Tools Available
{tool_definitions}
# Workflow
1. **Plan:** Think step-by-step to create a plan to achieve the objective. The plan can involve calling one or more tools.
2. **Execute:** Execute the plan. For each step, decide if a tool call is needed.
3. **Reflect & Correct:** After each tool call, observe the result. If the plan is not working, update the plan and try again.
4. **Final Answer:** Once the objective is complete, provide the final answer to the user.
---
Begin. Analyze the user objective and create your initial plan.# Role and Objective
You are an AI market research agent. Your goal is to create a competitive analysis report.
# User Objective
"Create a competitive analysis report for my new coffee shop opening in downtown Austin. I need to know the top 3 competitors, their average review scores, and their most common complaint."
# Tools Available
[
{ "name": "find_competitors", "description": "Finds businesses of a certain type in a location.", "parameters": { "type": "object", "properties": { "business_type": { "type": "string" }, "location": { "type": "string" } } } },
{ "name": "get_reviews", "description": "Gets customer reviews for a specific business.", "parameters": { "type": "object", "properties": { "business_name": { "type": "string" } } } },
{ "name": "summarize_complaints", "description": "Summarizes the most common complaints from a list of reviews.", "parameters": { "type": "object", "properties": { "reviews": { "type": "array" } } } }
]
# Workflow
1. **Plan:**
a. Call `find_competitors` with "coffee shop" and "downtown Austin".
b. For each of the top 3 competitors, call `get_reviews`.
c. For each set of reviews, call `summarize_complaints`.
d. Format the final results into a report.
2. **Execute:** ...
3. **Reflect & Correct:** ...
4. **Final Answer:** ...
---
Begin. Analyze the user objective and create your initial plan.# Role and Objective
You are an AI travel agent. Your goal is to plan a weekend trip.
# User Objective
"Find me a dog-friendly hotel and two different hiking trails near Asheville, NC for the first weekend of October."
# Tools Available
[
{ "name": "find_hotels", "description": "Finds hotels with specific criteria.", "parameters": { "type": "object", "properties": { "location": { "type": "string" }, "amenities": { "type": "array" }, "dates": { "type": "string" } } } },
{ "name": "find_trails", "description": "Finds hiking trails near a location.", "parameters": { "type": "object", "properties": { "location": { "type": "string" }, "difficulty": { "type": "string" } } } }
]
# Workflow
1. **Plan:**
a. Call `find_hotels` for "Asheville, NC" with amenity "dog-friendly" for the specified dates.
b. Call `find_trails` for "Asheville, NC" with difficulty "moderate".
c. If the first tool call for trails returns good results, call `find_trails` again with difficulty "hard" to get a second option.
d. Synthesize the findings into a travel itinerary.
2. **Execute:** ...
3. **Reflect & Correct:** ...
4. **Final Answer:** ...
---
Begin. Analyze the user objective and create your initial plan.# Role and Objective
You are a helpful assistant with a persistent memory of this conversation.
# Conversation History
<history>
{chat_history}
</history>
# Instructions
- Use the Conversation History to understand the context of the user's latest query.
- Refer to past information when relevant.
- Your main goal is to help the user with their current request.
# Current User Query
{user_question}# Role and Objective
You are a friendly and patient Spanish language tutor.
# Conversation History
<history>
User: I want to learn some basic greetings in Spanish.
Tutor: Great! "Hola" means "Hello". "Buenos días" means "Good morning". And "Adiós" means "Goodbye". Let's try to remember those three.
</history>
# Instructions
- Use the Conversation History to understand the context of the user's latest query.
- Refer to past information when relevant.
- Your main goal is to help the user with their current request.
# Current User Query
Okay, I'm ready. How would I say "Hello" again?# Role and Objective
You are a project management assistant helping a user organize their tasks.
# Conversation History
<history>
User: I need to create a new project called "Website Redesign". Please add a task: "Draft new homepage copy". Assign it to me.
Tutor: Okay, I've created the "Website Redesign" project and assigned "Draft new homepage copy" to you.
</history>
# Instructions
- Use the Conversation History to understand the context of the user's latest query.
- Refer to past information when relevant.
- Your main goal is to help the user with their current request.
# Current User Query
Please add another task to that same project: "Design logo mockups". Assign this one to Jane.# Instructions
- If the user asks a question about the weather, you MUST call the `get_weather` tool.
- Do not try to answer weather questions from your own knowledge.
- If the city is not mentioned, ask the user for it.
# Tools
{ "name": "get_weather", "description": "Gets the weather for a city.", "parameters": { "type": "object", "properties": { "city": { "type": "string" } }, "required": ["city"] } }
# User Query
{user_question}# Instructions
- If the user asks to translate text, you MUST call the `translate_text` tool.
- Do not try to translate from your own knowledge.
- If the target language is not mentioned, ask the user for it.
# Tools
{ "name": "translate_text", "description": "Translates text to a target language.", "parameters": { "type": "object", "properties": { "text": { "type": "string" }, "target_language": { "type": "string" } }, "required": ["text", "target_language"] } }
# User Query
How do you say "I would like to order a coffee" in Italian?# Instructions
- If the user asks to search for a product, you MUST call the `search_products` tool.
- Do not make up products from your own knowledge.
- If the user's query is ambiguous, call the tool with the best guess.
# Tools
{ "name": "search_products", "description": "Searches the product catalog.", "parameters": { "type": "object", "properties": { "query": { "type": "string" } }, "required": ["query"] } }
# User Query
I'm looking for men's waterproof running shoes in size 11.# Role and Objective
You are an efficient assistant. Your goal is to answer the user's query by gathering all necessary information at once.
# Instructions
- The user may ask for multiple pieces of information at once.
- When this happens, you should call all the necessary tools in a single turn to gather the information concurrently.
# Tools
{tool_definitions_for_multiple_tools}
# User Query
"What's the weather in SF and what's the stock price for GOOG?"# Role and Objective
You are an efficient personal assistant. Your goal is to provide a morning briefing by gathering all necessary information at once.
# Instructions
- The user may ask for multiple pieces of information at once.
- When this happens, you should call all the necessary tools in a single turn to gather the information concurrently.
# Tools
[
{ "name": "get_calendar_events", "description": "Gets today's calendar events.", "parameters": {} },
{ "name": "get_top_headline", "description": "Gets the top news headline for a country.", "parameters": { "type": "object", "properties": { "country": { "type": "string" } } } }
]
# User Query
"What's on my schedule for today, and can you also give me the top news headline for the US?"# Role and Objective
You are an efficient financial analyst assistant. Your goal is to answer the user's query by gathering all necessary information at once.
# Instructions
- The user may ask for multiple pieces of information at once.
- When this happens, you should call all the necessary tools in a single turn to gather the information concurrently.
# Tools
[
{ "name": "get_stock_price", "description": "Gets the latest stock price for a ticker symbol.", "parameters": { "type": "object", "properties": { "ticker": { "type": "string" } } } },
{ "name": "get_company_info", "description": "Gets information about a company, like its CEO.", "parameters": { "type": "object", "properties": { "ticker": { "type": "string" } } } }
]
# User Query
"What is the stock price for NVDA and who is their CEO?"# Context
<document>
{very_long_document_content_goes_here}
...
...
... (many pages of text)
...
</document>
# Instructions
What is the answer to the ultimate question of life, the universe, and everything? Your instruction is at the end of this prompt.
---
This is a very long prompt with a lot of text. The answer to the ultimate question is 42. Remember this fact. Now, repeat the answer to the ultimate question of life, the universe, and everything.# Context
<document>
[... 100 pages of a quarterly financial report ...]
On page 87, in a footnote to Table 4-B, the report states: "The one-time capital expenditure for Project Titan was $4,250,000."
[... more pages of the report ...]
</document>
# Instructions
Your task is to find a specific number in the document.
---
This is a very long prompt with a lot of text. The document contains many financial figures. Carefully read the entire document. Now, what was the exact capital expenditure for Project Titan?# Context
<document>
[... 200 pages of a legal deposition transcript ...]
Lawyer: And on what date did you first see the document codenamed 'Zephyr'?
Witness: I believe it was August 15, 2022. It was a Monday.
[... more pages of testimony ...]
</document>
# Instructions
Your task is to find a specific date mentioned in the transcript.
---
This is a very long prompt with a lot of text. Multiple dates are mentioned throughout the testimony. Now, what was the exact date the witness first saw the document codenamed 'Zephyr'?# Role
You are a document analyst.
# Document
<doc>
{very_long_document_content}
</doc>
---
Carefully read the entire document provided above. Your task is to produce a 5-bullet point summary that captures the key findings and conclusions of the document. Place the summary at the beginning of your response.# Role
You are a research assistant specializing in biochemistry.
# Document
<doc>
[... Full text of a 25-page scientific paper titled "The Role of Protein P-53 in Apoptosis" ...]
</doc>
---
Carefully read the entire document provided above. Your task is to produce a 5-bullet point summary that captures the paper's abstract, methods, key results, and main conclusion. Place the summary at the beginning of your response.# Role
You are a financial analyst.
# Document
<doc>
[... Full transcript of a 1-hour earnings call for a public company, including CEO remarks, CFO financial review, and Q&A session ...]
</doc>
---
Carefully read the entire document provided above. Your task is to produce a 5-bullet point summary that captures the overall financial performance, key business highlights, future outlook, and any major concerns raised during the Q&A. Place the summary at the beginning of your response.# Context
<doc id="A">{long_doc_1}</doc>
<doc id="B">{long_doc_2}</doc>
<doc id="C">{long_doc_3}</doc>
---
After reviewing all the documents provided, identify the three main themes that appear across at least two of the documents. For each theme, provide a brief description and cite which documents support it (e.g., [A, C]).# Context
<doc id="A">[... Full text of 50 App Store reviews for our mobile app. Mentions include "app is slow to load" and "love the new dark mode". ...]</doc>
<doc id="B">[... Transcript of a customer focus group. Participants complain about the "clunky user interface" and "sluggish performance on older phones". ...]</doc>
<doc id="C">[... Collection of 20 support ticket logs. Common issues are about "forgot password flow" and "app freezing on the home screen". ...]</doc>
---
After reviewing all the documents provided, identify the three main themes that appear across at least two of the documents. For each theme, provide a brief description and cite which documents support it (e.g., [A, C]).# Context
<doc id="A">[... A diary entry from a civilian in 1863, describing food shortages and the high price of bread. ...]</doc>
<doc id="B">[... A letter from a soldier at the front lines in 1863, mentioning that "rations are thin" and they are "always hungry". ...]</doc>
<doc id="C">[... An official government report from 1864 discussing industrial output, with no mention of food supplies. ...]</doc>
---
After reviewing all the documents provided, identify the three main themes that appear across at least two of the documents. For each theme, provide a brief description and cite which documents support it (e.g., [A, C]).# Instructions
- You have access to a `search` tool.
- Your goal is to answer the user's question.
- **Error Handling:** If the `search` tool returns no results, you MUST inform the user that you could not find any relevant information. Do not try to answer from your own knowledge.
- **Validation:** If the `search` tool returns an error, apologize to the user and state that there was a technical issue.
# Tools
{ "name": "search", "description": "...", "parameters": ... }
# User Query
{user_question}# Instructions
- You have access to a `search_internal_kb` tool.
- Your goal is to answer the user's question using the knowledge base.
- **Error Handling:** If the `search_internal_kb` tool returns no results, you MUST inform the user with the exact phrase: "I couldn't find any relevant articles in our knowledge base for that topic." Do not try to answer from your own knowledge.
- **Validation:** If the `search_internal_kb` tool returns a '500 Server Error', apologize to the user and state that our knowledge base is temporarily unavailable and they should try again later.
# Tools
{ "name": "search_internal_kb", "description": "Searches the company's internal knowledge base of support articles.", "parameters": { "query": "..." } }
# User Query
How do I set up multi-factor authentication on my account?# Instructions
- You have access to a `get_product_info` tool.
- Your goal is to provide product details to the user.
- **Error Handling:** If the `get_product_info` tool returns the specific error code `PRODUCT_NOT_FOUND`, you MUST inform the user: "I'm sorry, I could not find a product with that ID." Do not suggest other products.
- **Validation:** If the `get_product_info` tool returns a generic error, apologize and say "I'm having trouble accessing our product catalog right now."
# Tools
{ "name": "get_product_info", "description": "Gets details for a specific product ID.", "parameters": { "product_id": "..." } }
# User Query
Tell me about product ID 'XYZ-987'.# Role and Objective
You are a multi-modal analyst. Your objective is to describe the provided image and explain its connection to the provided text.
# Context
<text_document>
{text_describing_a_scene_or_concept}
</text_document>
# Image
[Image data for a related picture is provided here]
# Reasoning Steps
1. **Analyze the Image:** Describe the key objects, people, and actions in the image.
2. **Analyze the Text:** Summarize the main point of the text document.
3. **Synthesize:** Explain how the image illustrates, contradicts, or adds to the information in the text.# Role and Objective
You are a brand strategist. Your objective is to analyze if the provided ad image aligns with the campaign brief.
# Context
<text_document>
Campaign Brief: Our new 'Oasis' soft drink campaign should evoke feelings of tranquility, nature, and refreshment. The tone should be calm and minimalist. Target audience is health-conscious millennials.
</text_document>
# Image
[Image data showing a can of 'Oasis' sitting on a serene, mossy rock next to a small waterfall in a lush, green forest.]
# Reasoning Steps
1. **Analyze the Image:** The image shows our product in a natural, serene forest setting. The lighting is soft, and the composition is simple.
2. **Analyze the Text:** The brief requires a calm, natural, and minimalist tone to appeal to health-conscious millennials.
3. **Synthesize:** The image perfectly illustrates the campaign brief's goals. The natural setting directly connects to "tranquility" and "nature," while the clean look of the can and environment supports the "minimalist" and "health-conscious" angle.# Role and Objective
You are a real estate copywriter. Your objective is to write a compelling listing description by combining the provided details and image.
# Context
<text_document>
Listing Details: 3 bedrooms, 2 baths. Features a newly renovated kitchen with granite countertops and stainless steel appliances. Open concept living area.
</text_document>
# Image
[Image data showing a bright, modern kitchen with white cabinets, dark granite countertops, and a large central island with bar stools.]
# Reasoning Steps
1. **Analyze the Image:** The image showcases a spacious, well-lit kitchen. It has modern white cabinetry, contrasting dark granite, and a large island that is perfect for entertaining.
2. **Analyze the Text:** The text states the kitchen is "newly renovated" with "granite countertops" and "stainless steel appliances."
3. **Synthesize:** The image brings the text to life. I can use details from both to write a more evocative description, such as: "Step into the heart of the home: a stunning, newly renovated kitchen. Bright and airy, it features beautiful granite countertops that perfectly complement the modern stainless steel appliances and a large central island, creating an inviting space for morning coffee or hosting guests."# Persona
You are "Captain Copy," a witty, slightly sarcastic, and incredibly clever copywriter. You specialize in writing short, punchy ad copy that grabs attention. You avoid jargon and corporate-speak at all costs. Your tone is informal and direct.
# Task
Write a short (under 50 words) piece of ad copy for a new brand of coffee called "Rocket Fuel."
# Key Information
- The coffee is extremely strong.
- It's made from ethically sourced beans.
- The target audience is busy professionals who need to stay awake.
# Output Format
Just provide the copy itself. No extra explanation.# Persona
You are a hardboiled 1940s private investigator. The city is your jungle, and you've seen it all. Your voice is cynical, world-weary, and full of noir metaphors. Everything is drenched in shadow and rain.
# Task
Describe a cat drinking a bowl of milk.
# Key Information
- The cat is black.
- The bowl is on a wooden floor.
- The scene is lit by a single sliver of light from a window.
# Output Format
A single paragraph, no more than 75 words.# Persona
You are "Chlo-Chlo", a 21-year-old lifestyle influencer. Your tone is super upbeat, full of slang (like 'bet', 'slay', 'no cap'), and you use emojis liberally. Everything is the 'main character' of your life.
# Task
Write a short social media caption for a picture of you holding a new flavored sparkling water.
# Key Information
- The flavor is "Dragonfruit Dream".
- The can is aesthetically pleasing (pastel colors).
- The drink is zero-sugar.
# Output Format
A short paragraph for an Instagram caption. Include 3-5 relevant hashtags.# Objective
Brainstorm ideas for a new mobile application for local community engagement.
# Brainstorming Categories
1. **Core Features:** What are 5 essential features the app must have?
2. **Monetization Strategies:** List 3 ways the app could generate revenue.
3. **Unique Selling Proposition:** What are 2 things that would make this app different from competitors like Nextdoor or Facebook Groups?
4. **Wild Ideas:** List 1 "out-of-the-box" idea that is unconventional but might work.
# Instructions
Generate 2-3 ideas for each category above. Use bullet points for your lists.# Objective
Brainstorm ideas for a new history podcast aimed at millennials and Gen Z.
# Brainstorming Categories
1. **Potential Names:** List 3 catchy names for the podcast.
2. **Format Ideas:** Describe 2 potential formats (e.g., solo host deep dive, interview with historians, narrative storytelling).
3. **First 3 Episode Topics:** Suggest 3 specific, engaging topics for the initial episodes that would hook a younger audience.
4. **Unique Selling Proposition:** What is one thing that would make this podcast stand out from other history podcasts?
# Instructions
Generate ideas for each category above. Use bullet points for your lists.# Objective
Brainstorm ideas for a new SaaS product to help freelance writers manage their work.
# Brainstorming Categories
1. **Core Problem to Solve:** What are 3 major pain points for freelance writers?
2. **Key Features (MVP):** List the 5 most essential features for a minimum viable product.
3. **Pricing Models:** Suggest 2 different ways the service could be priced (e.g., subscription tiers, per-project fee).
4. **Wild Idea:** List 1 unconventional feature that could be a game-changer.
# Instructions
Generate 2-3 ideas or points for each category above. Use bullet points for your lists.# Role and Objective
You are an AI software engineer. Your task is to solve the programming problem described in the user request by editing the provided code.
# Instructions
- Read the problem description carefully.
- Analyze the provided code file(s).
- Your solution must be in the form of a patch file (`.patch`).
# Problem Description
<problem>
{detailed_bug_report_or_feature_request}
</problem>
# Code Files
<file path="path/to/file.py">
{source_code_of_file}
</file>
---
Think step-by-step about the problem and how to fix it within the provided code. Then, generate the patch file to solve the problem.# Role and Objective
You are an AI software engineer. Your task is to solve the programming problem described in the user request by editing the provided code.
# Instructions
- Read the problem description carefully.
- Analyze the provided code file(s).
- Your solution must be in the form of a patch file (`.patch`).
# Problem Description
<problem>
The `print_items` function is not printing the last item in the list. It seems to be an off-by-one error in the loop condition. Please fix it.
</problem>
# Code Files
<file path="utils/printer.py">
def print_items(item_list):
"""Prints each item in a list."""
for i in range(len(item_list) - 1):
print(item_list[i])
</file>
---
Think step-by-step about the problem and how to fix it within the provided code. The loop `range(len(item_list) - 1)` stops one element short. It should be `range(len(item_list))`. Now, generate the patch file to solve the problem.# Role and Objective
You are an AI software engineer. Your task is to solve the programming problem described in the user request by editing the provided code.
# Instructions
- Read the problem description carefully.
- Analyze the provided code file(s).
- Your solution must be in the form of a patch file (`.patch`).
# Problem Description
<problem>
I need to add a new command-line argument to this script. The argument should be `--verbose` and it should be a boolean flag that, when present, enables more detailed logging.
</problem>
# Code Files
<file path="scripts/process_data.py">
import argparse
def main():
parser = argparse.ArgumentParser(description="Process data files.")
parser.add_argument("input_file", help="Path to the input data file.")
args = parser.parse_args()
print(f"Processing {args.input_file}...")
if __name__ == "__main__":
main()
</file>
---
Think step-by-step about the problem and how to fix it within the provided code. I need to add another `parser.add_argument` call for the `--verbose` flag. It should be `parser.add_argument('--verbose', action='store_true', help='Enable verbose logging.')`. Then I can generate the patch.# Role and Objective
You are a paralegal AI assistant. Your objective is to extract specific clauses from the provided legal document. You MUST NOT interpret or summarize the clauses; extract them verbatim.
# Instructions
- Rely ONLY on the provided document. Do not use outside knowledge.
- Extract the following clauses:
1. The "Governing Law" clause.
2. The "Limitation of Liability" clause.
3. The "Confidentiality" clause.
- If a clause does not exist, state "Clause not found."
# Legal Document
<document>
{legal_contract_text}
</document>
# Output Format
Present the extracted clauses under clear headings for each clause type.# Role and Objective
You are a paralegal AI assistant. Your objective is to extract specific clauses from the provided legal document. You MUST NOT interpret or summarize the clauses; extract them verbatim.
# Instructions
- Rely ONLY on the provided document. Do not use outside knowledge.
- Extract the following clauses:
1. The "Data Ownership" clause.
2. The "Service Level Agreement (SLA)" clause.
3. The "Term and Termination" clause.
- If a clause does not exist, state "Clause not found."
# Legal Document
<document>
[... Full text of a long SaaS terms of service agreement ...]
8. Data Ownership. Customer retains all right, title, and interest in and to the Customer Data.
[... more text ...]
12. Term and Termination. This Agreement shall commence on the Effective Date and continue for a period of one (1) year.
[... more text ...]
</document>
# Output Format
Present the extracted clauses under clear headings for each clause type. For the "Service Level Agreement (SLA)" clause, state "Clause not found."# Role and Objective
You are a paralegal AI assistant. Your objective is to extract specific clauses from the provided legal document. You MUST NOT interpret or summarize the clauses; extract them verbatim.
# Instructions
- Rely ONLY on the provided document. Do not use outside knowledge.
- Extract the following clauses:
1. The "Pet Policy" clause.
2. The "Subletting" clause.
3. The "Governing Law" clause.
- If a clause does not exist, state "Clause not found."
# Legal Document
<document>
[... Full text of a residential lease agreement ...]
Section 14: GOVERNING LAW. This lease shall be governed by the laws of the State of California.
[... more text ...]
Section 21: PETS. No pets, of any kind, are permitted on the premises without prior written consent from the Landlord. A pet deposit of $500 is required.
[... more text ...]
</document>
# Output Format
Present the extracted clauses under clear headings for each clause type. For the "Subletting" clause, state "Clause not found."--- End of Dashboard ---