- Getting started
- Prerequisites
- Building agents in Studio Web
- Building agents in Agent Builder
Agents user guide
Agent score – Preview
The Agent score is an objective score used to identify areas where an agent needs optimization or additional training.
The score is based on a set of UiPath® Agents building best practices. For details, refer to the Best practices section.
-
Select Open health score from the right-side panel to view your agent's score.
Individual scores are available for each agent component: the prompt, the tools, the input schema, and the evaluation.
- Select a component to view a detailed breakdown of how the score was calculated.
Figure 1. How the agent score is calculated for the agent tools
-
For each component, select Help me improve this score to use Autopilot capabilities for suggestions.
-
Select Recalculate to run the analysis again.
The score considers the following elements:
- Prompt structure, purpose, length
- Tool, context, human in the loop usage
- Input/output arguments
- Evaluation sets
- Evaluation score
Each component is examined from multiple perspectives and combined into a single weighted score. The analysis runs in parallel, using LLMs (both completions and embeddings) with automatic retry mechanisms to handle rate limits.
For each component, you receive:
- A score from 0-100
- A detailed justification highlighting improvement areas
The final score report provides:
- A weighted average of all component scores
- Token usage and LLM call tracking
- An AI-generated summary explaining the overall score and key findings
