# Prompt Quality Assurance Checklist

This checklist ensures prompts meet quality standards and follow best practices for AI agent development. Use this for all prompt engineering work to maintain consistency and effectiveness.

## Prompt Design Fundamentals

### Clarity & Precision

- [ ] **Objective clearly stated** - Purpose and expected outcome explicitly defined
- [ ] **Ambiguity eliminated** - No unclear or vague language that could be misinterpreted
- [ ] **Context provided** - Sufficient background information included
- [ ] **Constraints specified** - Clear boundaries on what agent should/shouldn't do
- [ ] **Format requirements clear** - Expected output structure defined

### Structure & Organization

- [ ] **Logical flow maintained** - Instructions ordered logically from general to specific
- [ ] **Sections clearly delineated** - Different parts of prompt clearly separated
- [ ] **Role/persona defined** - Agent's role and expertise level specified
- [ ] **Instructions prioritized** - Most important requirements emphasized
- [ ] **Examples provided** - Where helpful, concrete examples included

### Language Quality

- [ ] **Grammar and spelling correct** - No errors that could cause confusion
- [ ] **Tone appropriate** - Matches intended interaction style
- [ ] **Terminology consistent** - Same concepts use same terms throughout
- [ ] **Concise yet complete** - No unnecessary words, but all essentials covered
- [ ] **Professional language** - Appropriate for business/professional context

## Prompt Engineering Best Practices

### Few-Shot Learning

- [ ] **Examples representative** - Cover key variations of expected inputs
- [ ] **Examples diverse** - Show different approaches and edge cases
- [ ] **Input-output pairs clear** - Relationship between example inputs and outputs obvious
- [ ] **Example quality high** - All examples demonstrate ideal responses
- [ ] **Number appropriate** - Not too few (incomplete) or too many (overwhelming)

### Chain of Thought

- [ ] **Reasoning steps included** - For complex tasks, thinking process shown
- [ ] **Logical progression** - Steps follow naturally from each other
- [ ] **Intermediate outputs defined** - What should be shown at each step
- [ ] **Error checking included** - How to verify correctness at each stage
- [ ] **Final synthesis clear** - How steps combine to reach conclusion

### Safety & Alignment

- [ ] **Harmful content prevention** - Instructions prevent inappropriate outputs
- [ ] **Bias mitigation addressed** - Prompts avoid reinforcing stereotypes
- [ ] **Privacy protection built-in** - Clear instructions about handling sensitive data
- [ ] **Ethical guidelines included** - Moral boundaries clearly established
- [ ] **Fallback behavior defined** - What to do when unsure or encountering edge cases

## Technical Implementation

### Token Efficiency

- [ ] **Length optimized** - No unnecessary verbosity that wastes tokens
- [ ] **Repetition minimized** - Same information not stated multiple times
- [ ] **Formatting efficient** - Uses minimal tokens for structure
- [ ] **Variable placeholders used** - Dynamic content properly templated
- [ ] **Cost impact assessed** - Token usage estimated and justified

### Modularity & Reusability

- [ ] **Components identified** - Reusable prompt sections clearly marked
- [ ] **Variables parameterized** - Dynamic elements properly abstracted
- [ ] **Templates created** - Common patterns made reusable
- [ ] **Inheritance structure** - Base prompts and specialized variants
- [ ] **Version control ready** - Prompt changes can be tracked

### Integration Readiness

- [ ] **API compatibility verified** - Works with target model APIs
- [ ] **Input validation considered** - How to handle malformed inputs
- [ ] **Output parsing planned** - How to extract structured data from responses
- [ ] **Error handling defined** - Graceful degradation when prompts fail
- [ ] **Timeout behavior specified** - What happens with slow responses

## Testing & Validation

### Functional Testing

- [ ] **Core use cases tested** - Primary functionality verified with real inputs
- [ ] **Edge cases covered** - Unusual, boundary, and extreme inputs tested
- [ ] **Error scenarios tested** - Behavior with invalid or problematic inputs
- [ ] **Cross-model compatibility** - Tested with different LLM models if applicable
- [ ] **Performance benchmarked** - Response time and quality measured

### Quality Assurance

- [ ] **Output consistency verified** - Similar inputs produce similar outputs
- [ ] **Accuracy validated** - Factual correctness checked where applicable
- [ ] **Completeness assessed** - Outputs include all required information
- [ ] **Relevance confirmed** - Responses address the actual question/task
- [ ] **Format compliance verified** - Outputs match specified structure

### Bias & Fairness Testing

- [ ] **Demographic variation tested** - Responses consistent across groups
- [ ] **Cultural sensitivity verified** - Appropriate for target audiences
- [ ] **Stereotype avoidance confirmed** - No reinforcement of harmful biases
- [ ] **Inclusive language used** - Welcoming to all user groups
- [ ] **Accessibility considered** - Usable by people with different needs

## Advanced Prompt Techniques

### Multi-Turn Conversations

- [ ] **Context management defined** - How conversation history is maintained
- [ ] **State tracking implemented** - Important information persisted across turns
- [ ] **Conversation flow planned** - Natural progression through interaction
- [ ] **Context window management** - How to handle token limits in long conversations
- [ ] **Reset conditions specified** - When and how to clear context

### Tool Integration

- [ ] **Function calling configured** - Available tools properly described
- [ ] **Parameter validation included** - How to validate tool inputs
- [ ] **Error handling for tools** - What to do when tools fail
- [ ] **Tool selection logic** - When to use which tools
- [ ] **Result interpretation** - How to process and present tool outputs

### Advanced Reasoning

- [ ] **Metacognitive prompts** - Agent reflects on its own reasoning
- [ ] **Uncertainty expression** - How to communicate confidence levels
- [ ] **Multiple perspectives** - Consideration of different viewpoints
- [ ] **Evidence evaluation** - How to assess information quality
- [ ] **Conclusion justification** - Why specific answers were chosen

## Performance Optimization

### Response Quality

- [ ] **Relevance maximized** - Responses directly address user needs
- [ ] **Depth appropriate** - Level of detail matches requirements
- [ ] **Accuracy prioritized** - Correctness over creativity when needed
- [ ] **Consistency maintained** - Similar queries get similar response quality
- [ ] **User satisfaction optimized** - Responses meet user expectations

### Efficiency Measures

- [ ] **Token usage minimized** - Efficient prompts without quality loss
- [ ] **Response time optimized** - Shorter prompts for time-sensitive tasks
- [ ] **Caching opportunities identified** - Parts of responses that can be reused
- [ ] **Batch processing considered** - Multiple requests handled efficiently
- [ ] **Model selection optimized** - Right model for task complexity

### Scalability Planning

- [ ] **Load testing completed** - Prompts tested under high volume
- [ ] **Resource requirements estimated** - Expected compute and cost
- [ ] **Bottlenecks identified** - Potential performance limitations
- [ ] **Optimization roadmap created** - Plan for future improvements
- [ ] **Monitoring metrics defined** - How to track prompt performance

## Documentation & Maintenance

### Documentation Requirements

- [ ] **Prompt purpose documented** - Why this prompt exists and what it does
- [ ] **Usage instructions clear** - How to implement and use the prompt
- [ ] **Example usage provided** - Sample inputs and expected outputs
- [ ] **Configuration options listed** - Available parameters and their effects
- [ ] **Troubleshooting guide included** - Common issues and solutions

### Version Control

- [ ] **Change history tracked** - All modifications documented
- [ ] **Version numbers assigned** - Clear versioning scheme followed
- [ ] **Backward compatibility considered** - Impact on existing implementations
- [ ] **Migration guide provided** - How to upgrade from previous versions
- [ ] **Rollback procedure defined** - How to revert if issues arise

### Maintenance Planning

- [ ] **Update schedule defined** - Regular review and improvement cycle
- [ ] **Performance monitoring enabled** - Ongoing quality assessment
- [ ] **Feedback collection planned** - How to gather user input
- [ ] **Improvement process established** - How to implement enhancements
- [ ] **Deprecation policy clear** - How old prompts are retired

## Security & Compliance

### Security Considerations

- [ ] **Injection resistance tested** - Prompt injection attacks prevented
- [ ] **Information leakage prevented** - No unintended data exposure
- [ ] **Access controls considered** - Who can modify prompts
- [ ] **Audit trail maintained** - Changes tracked for security
- [ ] **Vulnerability assessment done** - Security review completed

### Compliance Requirements

- [ ] **Regulatory compliance verified** - Meets industry standards
- [ ] **Privacy requirements met** - Data protection regulations followed
- [ ] **Content policies followed** - Platform and organizational guidelines
- [ ] **Legal review completed** - Attorney approval where required
- [ ] **Compliance documentation updated** - Records maintained properly

## Review & Approval

### Technical Review

- [ ] **Peer review completed** - Another prompt engineer reviewed
- [ ] **Technical accuracy verified** - Subject matter expert validated
- [ ] **Implementation tested** - Actually deployed and verified
- [ ] **Performance acceptable** - Meets speed and quality requirements
- [ ] **Integration successful** - Works with existing systems

### Business Review

- [ ] **Requirements met** - Fulfills business objectives
- [ ] **User experience approved** - UX team validated interaction
- [ ] **Cost justified** - Token usage within budget
- [ ] **Risk assessment done** - Potential issues identified and mitigated
- [ ] **Stakeholder approval obtained** - Business owner signed off

## Sign-offs

### Technical Approval

- [ ] **Prompt Engineer**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**
- [ ] **Technical Lead**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**
- [ ] **QA Engineer**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**
- [ ] **Security Review**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**

### Business Approval

- [ ] **Product Owner**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**
- [ ] **UX Designer**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**
- [ ] **Compliance Officer**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**
- [ ] **Business Stakeholder**: \***\*\*\*\*\*\*\***\_\***\*\*\*\*\*\*\*** Date: **\_\_\_**

## Quality Metrics

Document the following metrics for this prompt:

- **Average Response Quality Score**: \_\_\_\_/10
- **Token Efficiency Score**: \_\_\_\_/10 (quality/tokens ratio)
- **User Satisfaction Rating**: \_\_\_\_/10
- **Accuracy Rate**: \_\_\_\_%
- **Consistency Score**: \_\_\_\_/10
- **Safety Score**: \_\_\_\_/10

## Notes and Improvements

### Known Limitations

1. ***
2. ***
3. ***

### Future Improvements

1. ***
2. ***
3. ***

### Additional Comments

_Space for specific notes about this prompt:_

---

---

---

---

**Quality Approval**: [ ] APPROVED [ ] NEEDS REVISION [ ] REJECTED

**Date**: **\*\***\_\_\_\_**\*\***

**Final Approver**: **\*\***\*\***\*\***\_\_\_\_**\*\***\*\***\*\***