Prompt Engineering

1---
2name: prompt-engineering
3description: Expert guide on prompt engineering patterns, best practices, and optimization techniques. Use when user wants to improve prompts, learn prompting strategies, or debug agent behavior.
4---
5 
6# Prompt Engineering Patterns
7 
8Advanced prompt engineering techniques to maximize LLM performance, reliability, and controllability.
9 
10## Core Capabilities
11 
12### 1. Few-Shot Learning
13 
14Teach the model by showing examples instead of explaining rules. Include 2-5 input-output pairs that demonstrate the desired behavior. Use when you need consistent formatting, specific reasoning patterns, or handling of edge cases. More examples improve accuracy but consume tokens—balance based on task complexity.
15 
16**Example:**
17 
18```markdown
19Extract key information from support tickets:
20 
21Input: "My login doesn't work and I keep getting error 403"
22Output: {"issue": "authentication", "error_code": "403", "priority": "high"}
23 
24Input: "Feature request: add dark mode to settings"
25Output: {"issue": "feature_request", "error_code": null, "priority": "low"}
26 
27Now process: "Can't upload files larger than 10MB, getting timeout"
28```
29 
30### 2. Chain-of-Thought Prompting
31 
32Request step-by-step reasoning before the final answer. Add "Let's think step by step" (zero-shot) or include example reasoning traces (few-shot). Use for complex problems requiring multi-step logic, mathematical reasoning, or when you need to verify the model's thought process. Improves accuracy on analytical tasks by 30-50%.
33 
34**Example:**
35 
36```markdown
37Analyze this bug report and determine root cause.
38 
39Think step by step:
40 
411. What is the expected behavior?
422. What is the actual behavior?
433. What changed recently that could cause this?
444. What components are involved?
455. What is the most likely root cause?
46 
47Bug: "Users can't save drafts after the cache update deployed yesterday"
48```
49 
50### 3. Prompt Optimization
51 
52Systematically improve prompts through testing and refinement. Start simple, measure performance (accuracy, consistency, token usage), then iterate. Test on diverse inputs including edge cases. Use A/B testing to compare variations. Critical for production prompts where consistency and cost matter.
53 
54**Example:**
55 
56```markdown
57Version 1 (Simple): "Summarize this article"
58→ Result: Inconsistent length, misses key points
59 
60Version 2 (Add constraints): "Summarize in 3 bullet points"
61→ Result: Better structure, but still misses nuance
62 
63Version 3 (Add reasoning): "Identify the 3 main findings, then summarize each"
64→ Result: Consistent, accurate, captures key information
65```
66 
67### 4. Template Systems
68 
69Build reusable prompt structures with variables, conditional sections, and modular components. Use for multi-turn conversations, role-based interactions, or when the same pattern applies to different inputs. Reduces duplication and ensures consistency across similar tasks.
70 
71**Example:**
72 
73```python
74# Reusable code review template
75template = """
76Review this {language} code for {focus_area}.
77 
78Code:
79{code_block}
80 
81Provide feedback on:
82{checklist}
83"""
84 
85# Usage
86prompt = template.format(
87    language="Python",
88    focus_area="security vulnerabilities",
89    code_block=user_code,
90    checklist="1. SQL injection\n2. XSS risks\n3. Authentication"
91)
92```
93 
94### 5. System Prompt Design
95 
96Set global behavior and constraints that persist across the conversation. Define the model's role, expertise level, output format, and safety guidelines. Use system prompts for stable instructions that shouldn't change turn-to-turn, freeing up user message tokens for variable content.
97 
98**Example:**
99 
100```markdown
101System: You are a senior backend engineer specializing in API design.
102 
103Rules:
104 
105- Always consider scalability and performance
106- Suggest RESTful patterns by default
107- Flag security concerns immediately
108- Provide code examples in Python
109- Use early return pattern
110 
111Format responses as:
112 
1131. Analysis
1142. Recommendation
1153. Code example
1164. Trade-offs
117```
118 
119## Key Patterns
120 
121### Progressive Disclosure
122 
123Start with simple prompts, add complexity only when needed:
124 
1251. **Level 1**: Direct instruction
126 
127   - "Summarize this article"
128 
1292. **Level 2**: Add constraints
130 
131   - "Summarize this article in 3 bullet points, focusing on key findings"
132 
1333. **Level 3**: Add reasoning
134 
135   - "Read this article, identify the main findings, then summarize in 3 bullet points"
136 
1374. **Level 4**: Add examples
138   - Include 2-3 example summaries with input-output pairs
139 
140### Instruction Hierarchy
141 
142```
143[System Context] → [Task Instruction] → [Examples] → [Input Data] → [Output Format]
144```
145 
146### Error Recovery
147 
148Build prompts that gracefully handle failures:
149 
150- Include fallback instructions
151- Request confidence scores
152- Ask for alternative interpretations when uncertain
153- Specify how to indicate missing information
154 
155## Best Practices
156 
1571. **Be Specific**: Vague prompts produce inconsistent results
1582. **Show, Don't Tell**: Examples are more effective than descriptions
1593. **Test Extensively**: Evaluate on diverse, representative inputs
1604. **Iterate Rapidly**: Small changes can have large impacts
1615. **Monitor Performance**: Track metrics in production
1626. **Version Control**: Treat prompts as code with proper versioning
1637. **Document Intent**: Explain why prompts are structured as they are
164 
165## Common Pitfalls
166 
167- **Over-engineering**: Starting with complex prompts before trying simple ones
168- **Example pollution**: Using examples that don't match the target task
169- **Context overflow**: Exceeding token limits with excessive examples
170- **Ambiguous instructions**: Leaving room for multiple interpretations
171- **Ignoring edge cases**: Not testing on unusual or boundary inputs
172