AI Quality Assurance Agent for Outputs

# AI Quality Assurance Agent for Outputs

You are a software quality architect designing an AI QA agent for [DEVELOPMENT_TEAM] building [PRODUCT_TYPE].

## Quality Evaluation Framework

Define evaluation dimensions for [OUTPUT_TYPE] outputs:

**Dimension 1 — Accuracy:**
Criteria: [ACCURACY_CRITERIA]
Scoring method (0–100):
Failure threshold (score below which output is rejected):

**Dimension 2 — Completeness:**
Criteria: [COMPLETENESS_CRITERIA]
Scoring and threshold:

**Dimension 3 — Format Compliance:**
Criteria: [FORMAT_CRITERIA]
Scoring and threshold:

**Dimension 4 — Safety and Policy:**
Criteria: [SAFETY_CRITERIA]
Hard block triggers (any of these = immediate rejection):

## Evaluator Agent Prompt

Write the complete evaluator prompt that scores [OUTPUT_TYPE]:

"Given the following output: [OUTPUT]
Expected criteria: [EXPECTED_CRITERIA]
Score each dimension 0–100. Flag hard failures immediately. For any score below 80, provide specific feedback the generating agent can act on."

## Feedback Loop Design

Define how failed outputs are handled:
1. Rejection → correction instructions sent back to generating agent
2. Pattern tracking → which failure types are increasing
3. Threshold alert → when failure rate exceeds [THRESHOLD]
4. Weekly quality report for [DEVELOPMENT_TEAM]

## Human Review Queue

Design the borderline review interface for outputs scoring 60–79 across dimensions.

## Continuous Improvement

Write the monthly analysis prompt that reviews QA data and recommends updates to evaluation criteria.