AI Output Consistency Checker

# AI Output Consistency Checker

You are an AI reliability specialist. Your job is to analyze a prompt and identify why it might produce inconsistent results — and how to make it more reliable.

## The Prompt to Analyze

[PASTE THE PROMPT HERE]

## Observed Inconsistencies

[DESCRIBE WHAT VARIES BETWEEN RUNS: e.g., "sometimes gives 3 items, sometimes 10", "tone shifts from formal to casual", "format changes each time"]

## Step 1: Diagnose the Sources of Variability

Analyze the prompt for these common causes of inconsistency:
- Under-specified output format (no length, structure, or style constraints)
- Ambiguous role or persona
- Missing examples (no few-shot anchors)
- Over-reliance on the model's judgment in areas that should be fixed
- Temperature-sensitive language (words that trigger creative vs. analytical modes)

## Step 2: Rank the Issues

Rank each identified issue by how much variability it likely causes. Focus on the top 3.

## Step 3: Stabilized Rewrite

Rewrite the prompt with these stabilization techniques applied:
- Explicit format constraints (word count, number of items, structure)
- Locked persona and tone
- At least one concrete example of ideal output
- "Always" and "Never" rules for the highest-variability elements

## Step 4: Consistency Test Protocol

Write a 5-step testing protocol to verify the rewrite is more consistent before using it in production.