Feedback: lemur-improving-your-prompt

Documentation Feedback

Original URL: https://www.assemblyai.com/docs/lemur/improving-your-prompt
Category: lemur
Generated: 05/08/2025, 4:29:04 pm

---

Claude Sonnet 4 Feedback

Generated: 05/08/2025, 4:29:03 pm

Documentation Analysis: Prompt Engineering for LeMUR

Overall Assessment

This documentation provides a solid foundation for prompt engineering with LeMUR, but there are several opportunities to improve clarity, completeness, and user experience.

Specific Issues & Recommendations

1. Missing Critical Information

Problem: Key technical details are absent

• No information about prompt length limits
• Missing details about LeMUR-specific features or limitations
• No error handling guidance
• Lack of performance considerations

Fix: Add a “Technical Specifications” section:

## Technical Specifications
- Maximum prompt length: [X] characters
- Supported output formats: JSON, Markdown, plain text
- Processing time: Typically 10-30 seconds depending on transcript length
- Rate limits: [X] requests per minute

2. Weak Examples Throughout

Problem: Examples are too generic and don’t demonstrate real-world complexity

Current example:

Provide a summary of the meeting.

Improved example:

Create a structured summary of this sales call focusing on client needs and next steps.

Context:
This is a discovery call with a potential enterprise client in the healthcare industry. Focus on compliance requirements, technical concerns, and budget discussions.

Answer Format:
## Client Overview
**Company:** <company_name>
**Industry:** <industry>
**Key Stakeholders:** <list_of_participants>

## Key Requirements
- <requirement_1>
- <requirement_2>

## Next Steps
- <action_item_with_owner_and_date>

3. Poor Structure and Flow

Problem: Information is scattered and lacks logical progression

Recommended restructure:

1. Quick Start (simple example that works immediately)
2. Core Concepts (anatomy of a prompt)
3. Best Practices (current instruction/context/format sections)
4. Advanced Techniques (complex examples, troubleshooting)
5. Reference (limits, specifications, common patterns)

4. Unclear Explanations

Problem: The “Context” section explanation is confusing

Current: “Add a Context: section to your prompt to provide additional information to the LLM.”

Improved:

The Context section helps LeMUR understand the specific domain, terminology, or requirements for your task. Since LeMUR automatically includes your transcript, use Context to add:
- Industry-specific definitions
- Expected output style/tone
- Relevant examples from similar tasks
- Special requirements or constraints

5. Major User Pain Points

A. No Troubleshooting Section

Add:

## Troubleshooting Common Issues

### Generic or Irrelevant Responses
**Symptoms**: LeMUR provides vague summaries or misses key points
**Solutions**:
- Add specific examples in your Context section
- Use more precise instruction verbs (extract, categorize, analyze vs. "tell me about")
- Break complex requests into smaller, focused prompts

### Inconsistent Formatting
**Symptoms**: Output doesn't match your specified format
**Solutions**:
- Provide complete format examples, not just templates
- Use "Answer Format:" exactly as shown
- Add "Follow this format exactly" to your instruction

B. No Progressive Learning Path

Add a “Getting Started” section:

## Getting Started: Your First Prompt

1. **Start Simple**: Begin with basic instructions

List the main topics discussed in this meeting.

2. **Add Context**: Include relevant background

List the main topics discussed in this meeting.

Context: This is a weekly engineering standup meeting where team members discuss progress and blockers.

3. **Specify Format**: Define your desired output structure
[Complete example here]

6. Missing Practical Guidance

Add sections for:

• Common Use Cases: Meeting summaries, action items, sentiment analysis, Q&A extraction
• Industry-Specific Examples: Sales calls, interviews, support calls, podcasts
• Integration Patterns: How prompts fit into workflows

7. Improve Code Examples

Current JSON example is incomplete:

{
  "assignee": <assignee>,
  "action_item": <action item>,
  "due_date": <due_date>
}

Should include expected output:

// Input format for Answer Format section:
[
  {
    "assignee": "<person_name>",
    "action_item": "<specific_task>",
    "due_date": "<YYYY-MM-DD or 'Not specified'>"
  }
]

// Example output:
[
  {
    "assignee": "Sarah Chen",
    "action_item": "Send revised proposal to client with updated pricing",
    "due_date": "2024-01-15"
  }
]

8. Add Interactive Elements

Recommendations:

• Add a “Prompt Builder” tool or template
• Include “Copy to clipboard” buttons for examples
• Add expandable sections for complex examples

9. Better Cross-References

Current: Vague reference to “LeMUR Examples” Improved: Specific links like “See our Meeting Summary Templates for industry-specific examples”

Priority Implementation Order

1. High Priority: Add troubleshooting section and improve examples
2. Medium Priority: Restructure content flow and add technical specifications
3. Low Priority: Add interactive elements and advanced techniques

These changes would significantly improve user success rates and reduce support requests while making the documentation more actionable and user-friendly.

---