Prompt Analysis

Overview

You CAN analyze prompts for quality and best practices by asking me for my prompt, then running the analysis script via Bash tool. The script returns AI-powered feedback categorized as good/warning/critical that you should explain to me.

When to Use

Use this skill when I:

• Ask you to analyze or review a prompt I am working on
• Say my prompts aren't giving good results and wants help improving them
• Want to learn prompt engineering best practices
• Is unsure if my prompt is clear or specific enough
• Ask "Can you help me write a better prompt?"

Skip when:

• My prompt is already clear and well-structured
• The conversation is casual (not about prompt engineering)
• I have not indicated I want prompt analysis help

Step-by-Step Instructions

1. Ask Me for My Prompt

Explicitly request the prompt to analyze:

I can analyze that prompt for you. Please share the exact prompt text you'd like me to review, and I'll run it through the analysis tool to identify areas for improvement.

Wait for me to provide the prompt. Do not proceed to analysis without receiving it.

2. Run the Analysis Script

Once you have the prompt, use the Bash tool to run the analysis:

node ~/.claude/skills/prompt-analysis/script.js --prompt="<my prompt here>"

Important:

• Escape the prompt properly for bash (use single quotes or escape double quotes)
• If the prompt contains single quotes, escape them: '\''
• If the prompt is very long, you may need to use a different quoting strategy

3. Read and Interpret the Output

The script returns colorized text output with three categories:

• Good (green ✓): Things I am doing well
• Warning (yellow ⚠): Areas that could be improved
• Critical (red ✗): Issues that should be addressed immediately

4. Explain the Feedback to Me

Don't just paste the raw output. Instead:

1. Summarize the overall quality - Start with a high-level assessment
2. Explain critical issues first - These are blockers that must be fixed
3. Address warnings next - These would improve results significantly
4. Acknowledge what's good - Positive reinforcement for what I am doing well
5. Provide specific examples - Show concrete improvements I can make
6. Offer to analyze a revised version - Encourage iteration

5. Optional: Iterate

If I revise my prompt based on your feedback, offer to run the analysis again:

Would you like me to analyze your revised prompt to see if we've addressed those issues?

Example Workflow

Me: "Can you help me improve this prompt: Write a function to sort numbers"

You: I can analyze that prompt for you. Let me run it through the analysis tool.

[Run: node ~/.claude/skills/prompt-analysis/script.js --prompt="Write a function to sort numbers"]

[Output shows warnings about missing language, algorithm, format details]

You: I've analyzed your prompt. Here's what the analysis found:

**Critical Issues:**
The prompt is quite vague - it doesn't specify which programming language you want, what sorting algorithm to use, or what format the input/output should be.

**Areas for Improvement:**
- Specify the programming language (JavaScript? Python? Java?)
- Define the input format (array? list? arguments?)
- Clarify the expected output (sorted array? in-place sort?)
- Mention edge cases you care about (empty arrays? duplicates?)

**What's Working:**
The request is concise and clear about the core action (sorting).

**Here's a revised version:**
"Write a JavaScript function called sortNumbers that takes an array of numbers as input and returns a new array sorted in ascending order. Handle edge cases like empty arrays and arrays with duplicates."

Would you like me to analyze this revised version?

Common Mistakes

❌ Running analysis without asking for the prompt first

Reality: Always ask me to provide my prompt explicitly before running analysis.

❌ Pasting raw script output without explanation

Reality: The output is for YOUR interpretation - explain it in plain language with specific examples.

❌ Skipping the "what's good" feedback

Reality: Always acknowledge positive aspects before diving into criticism.

❌ Not offering to iterate

Reality: Prompt improvement is iterative - offer to analyze revised versions.

Requirements

• Paid Nori subscription
• Configured credentials in ~/nori-config.json

Technical Details

The script calls the /api/prompt-analysis endpoint which uses an LLM to analyze the prompt and return structured feedback. The analysis considers:

• Clarity and specificity
• Completeness of requirements
• Context and constraints
• Examples and edge cases
• Tone and structure