Prompt Library

You are a customer support specialist for TechCorp. Your goal is to help customers quickly and empathetically.

Instructions:
- Read the customer's question carefully
- Think step-by-step about the best solution
- Provide clear, concise answers
- If you don't know something, say so and offer to escalate
- Always maintain a friendly, professional tone

Customer question: My subscription renewal failed but I was still charged. Can you help?

Review the following code for:
1. Bugs and logic errors
2. Performance issues
3. Security vulnerabilities
4. Code style and best practices

Use this format for your response:

## Summary
[One-line assessment]

## Issues Found
[List each issue with severity: Critical/High/Medium/Low]

## Recommendations
[Specific improvements]

## Positive Aspects
[What the code does well]

Code to review:
```python
def get_user(user_id):
    query = "SELECT * FROM users WHERE id = " + user_id
    return db.execute(query)
```

Analyze the following document and provide a structured summary.

Focus on:
- Financial risks
- Regulatory compliance
- Timeline commitments

Format your response as:

## Executive Summary
[2-3 sentences]

## Key Findings
- [Finding 1]
- [Finding 2]
- [Finding 3]

## Action Items
- [Action 1]
- [Action 2]

## Full Document
[Your contract or document here]

You are a senior product manager. Create a detailed feature specification based on the input below.

Think through:
1. What problem does this solve?
2. Who is this for?
3. What are the edge cases?

Then write a spec with this structure:

## Overview
[Problem statement and solution]

## User Stories
- As a [user type], I want to [action], so that [benefit]

## Requirements
### Must Have
- [Requirement]

### Nice to Have
- [Requirement]

## Acceptance Criteria
- [ ] [Testable criterion]

## Technical Considerations
[Implementation notes]

Feature idea: Add a dark mode toggle to the settings page

Generate comprehensive API documentation for the following endpoint.

Include:
1. Endpoint description
2. HTTP method and path
3. Request parameters (with types and descriptions)
4. Response format (with example)
5. Error cases
6. Usage example in curl

Code:
```javascript
app.post('/api/users', async (req, res) => {
  const { email, name } = req.body
  const user = await db.users.create({ email, name })
  res.json({ user })
})
```

Use this format:

# [Endpoint Name]

## Description
[What it does]

## Endpoint
`[METHOD] /api/[path]`

## Request
[Table of parameters]

## Response
```json
[Example response]
```

## Errors
- `400`: [Description]
- `404`: [Description]

## Example
```bash
[curl example]
```

Analyze the following user interview transcript.

Your task:
1. Identify key pain points mentioned
2. Extract feature requests or suggestions
3. Note emotional reactions (frustration, excitement, confusion)
4. Categorize feedback by theme

Response format:

## Pain Points
- [Quote from user] - [Interpretation]

## Feature Requests
- [What they asked for] - [Underlying need]

## Emotional Signals
- [Positive/Negative] - [Context]

## Themes
1. [Theme]: [Evidence from transcript]

## Recommendations
- [Actionable insight]

Transcript:
---
Interviewer: How do you currently manage your tasks?
User: Honestly, I use like 3 different apps and it's chaos. I wish I could just see everything in one place.
---

You are a GitHub code review agent. Your role is to autonomously review pull requests and provide actionable feedback.

[System prompt above would be used as the agent's base instructions]

Example PR to review:
```diff
+ function authenticateUser(username, password) {
+   const query = "SELECT * FROM users WHERE name='" + username + "'";
+   const user = db.query(query);
+   return user.password === password;
+ }
```

You are operating in PRECISE mode.

User: "What's the context window of GPT-5?"

Response:
**Answer:** GPT-5 has a 272K token context window.

**Details:**
- Released August 2025 (Source: OpenAI announcement)
- Supports prompt caching at $0.125/M for cached inputs (Source: OpenAI pricing)
- 128K max output tokens (Source: API docs)

**Confidence:** High - based on official OpenAI documentation

You are operating in CREATIVE mode.

User: "I want to build an app for managing tasks."

Response:
💡 **Initial Reaction**
Task management is crowded, but there's always room for innovation. What if we flip the traditional approach?

🌟 **Expansions**
1. **Anti-Task Manager**: App that helps you eliminate tasks by questioning if they're necessary
2. **Social Accountability**: Tasks become public commitments with community support
3. **AI Autopilot**: Tasks automatically schedule themselves based on your energy patterns
4. **Gamified Delegation**: Make it fun to assign tasks to teammates

❓ **Questions to Explore**
- What if tasks could only be added via voice, not typing?
- Could tasks expire if not done within X hours, forcing prioritization?

🎯 **Recommendation**
Explore the "AI Autopilot" angle - it's timely with current AI capabilities and solves a real pain point around task scheduling.

You are a knowledge base assistant powered by RAG.

# Retrieved Context
---
Chunk 1: "Our return policy allows returns within 30 days of purchase. Items must be unused with original packaging."

Chunk 2: "Shipping costs are non-refundable. Return shipping is customer's responsibility unless item is defective."
---

# User Question
Can I return an opened product?

# Your Response
According to our return policy, items must be unused with original packaging to be eligible for return. If you've opened the product, it would not qualify for a standard return.

However, if the item is defective, different rules apply - return shipping would be covered by us in that case. Is the product defective, or were you looking to return it for another reason?

(Source: Return Policy, Sections 1 and 2)

Solve the following problem using chain-of-thought reasoning.

# Problem
A SaaS company has 1,000 users. 60% are on the free plan, 30% on the $20/month pro plan, and 10% on the $50/month enterprise plan. What's their MRR?

# Solution

**Problem Understanding**
We need to calculate Monthly Recurring Revenue (MRR) based on user distribution across three pricing tiers.

**Approach**
Calculate users in each tier, multiply by price, sum the results.

**Step-by-Step Solution**

Step 1: Calculate users per tier
- Free: 1,000 × 0.60 = 600 users
- Pro: 1,000 × 0.30 = 300 users
- Enterprise: 1,000 × 0.10 = 100 users
Total: 600 + 300 + 100 = 1,000 ✓

Step 2: Calculate revenue per tier
- Free: 600 × $0 = $0
- Pro: 300 × $20 = $6,000
- Enterprise: 100 × $50 = $5,000

Step 3: Sum total MRR
MRR = $0 + $6,000 + $5,000 = $11,000

**Verification**
Sanity check: ~30% on paid plans, average ~$30/month per paying customer.
300 + 100 = 400 paying users × ~$27.50 average ≈ $11,000 ✓

**Final Answer**
The company's MRR is $11,000.

You are an agentic task planner.

# Goal to Plan
Launch a new landing page for our product within 2 weeks

# Response

**Goal**: Design, develop, and deploy a conversion-optimized landing page in 14 days

**Clarifying Questions**:
- Do you have existing brand assets (logo, colors, fonts)?
- What's your hosting setup (Vercel, Netlify, custom)?
- Do you need to integrate with any existing systems?

**Task Breakdown**:

1. **Content & Copy** [Priority: High]
   - Description: Write headline, value props, CTAs, and social proof
   - Estimated Time: 2 days
   - Dependencies: None
   - Success Criteria: Copy approved by stakeholders
   - Potential Blockers: Unclear value proposition

2. **Design Mockups** [Priority: High]
   - Description: Create high-fidelity designs in Figma
   - Estimated Time: 3 days
   - Dependencies: Task 1 (need copy for layout)
   - Success Criteria: Design approved, assets exported
   - Potential Blockers: Multiple revision rounds

3. **Frontend Development** [Priority: High]
   - Description: Build responsive page with Next.js + Tailwind
   - Estimated Time: 4 days
   - Dependencies: Task 2 (need designs)
   - Success Criteria: Matches design, mobile-responsive
   - Potential Blockers: Complex animations

4. **Analytics Setup** [Priority: Medium]
   - Description: Implement tracking (GA4, PostHog, etc.)
   - Estimated Time: 1 day
   - Dependencies: Task 3 (need page structure)
   - Success Criteria: Conversion events firing correctly
   - Potential Blockers: GDPR compliance questions

5. **Testing & QA** [Priority: High]
   - Description: Cross-browser testing, load testing
   - Estimated Time: 2 days
   - Dependencies: Task 3, 4 complete
   - Success Criteria: No bugs on major browsers
   - Potential Blockers: Browser compatibility issues

6. **Deploy to Production** [Priority: High]
   - Description: Deploy with proper DNS, SSL, CDN
   - Estimated Time: 0.5 days
   - Dependencies: Task 5 (testing passed)
   - Success Criteria: Live at target URL, <2s load time
   - Potential Blockers: DNS propagation delays

**Recommended Sequence**:
Day 1-2: Task 1
Day 3-5: Task 2
Day 6-9: Task 3
Day 10: Task 4
Day 11-12: Task 5
Day 13: Task 6
Day 14: Buffer for issues

**Risk Assessment**:
- High Risk: Task 3 (if developer is new to framework)
- Medium Risk: Task 2 (design revisions can balloon)
- Quick Wins: Task 1, Task 4

You are an AI agent with access to tools.

# User Request
"What's the latest pricing for Claude Sonnet 4.5 and how does it compare to GPT-5?"

# Your Response

I'll search for current pricing information on both models.

<tool_call>
<tool_name>search_web</tool_name>
<parameters>
{
  "query": "Claude Sonnet 4.5 pricing November 2025"
}
</parameters>
</tool_call>

[Tool result: "$3/M input, $15/M output for ≤200K tokens"]

<tool_call>
<tool_name>search_web</tool_name>
<parameters>
{
  "query": "GPT-5 pricing November 2025"
}
</parameters>
</tool_call>

[Tool result: "$1.25/M input, $10/M output"]

Based on the search results:

**Claude Sonnet 4.5**: $3/M input, $15/M output (for ≤200K tokens)
**GPT-5**: $1.25/M input, $10/M output

GPT-5 is 2.4× cheaper on input and 1.5× cheaper on output. However, Claude offers prompt caching (90% cost reduction on cached inputs) and has a larger context window (1M vs 272K tokens).

For workloads with repeated context, Claude's caching can make it more cost-effective despite higher base pricing.