Does it support live testing with real data?

Yes, it supports a live preview mode that uses real Flows and Apex actions, though a Connected App setup is required for OAuth.

Can this skill automatically fix my failing agents?

Yes, through the Agentic Fix Loop, it analyzes test failures and delegates fixes to the sf-ai-agentforce skill to update topic and action descriptions.

What are the requirements for using this skill?

It requires Salesforce API v65.0+ (Winter '26) and the Agent Testing Center feature enabled in your scratch org or Salesforce instance.

How is the testing coverage calculated?

The skill tracks Topic Selection accuracy and Action Invocation rates, providing a percentage-based summary of how much of your agent's logic was exercised.

Salesforce Agentforce Testing

Name: Salesforce Agentforce Testing
Author: Jaganpro

byJaganpro

•

Security & Testing

Automates Agentforce agent validation, coverage analysis, and autonomous fix loops for Salesforce AI agents.

About

This skill provides a robust framework for testing Salesforce Agentforce agents directly within Claude Code. It enables developers to execute automated tests via the sf CLI, generate comprehensive YAML test specifications, and perform deep analysis of topic and action coverage. By implementing a unique 'Agentic Fix Loop,' the skill can automatically detect failures—such as topic selection mismatches or action invocation errors—and coordinate with other Salesforce skills to repair the agent metadata, ensuring high-performance AI implementations that meet production standards.

Key Features

Comprehensive 100-point scoring system across five critical validation categories
YAML-based test spec generation for standardized and repeatable agent testing
32 GitHub stars
Interactive simulated and live preview modes for debugging agent behavior
Automated test execution with real-time coverage metrics for topics and actions
Agentic Fix Loop for autonomous resolution of failing test cases

Use Cases

Verifying agent routing accuracy and action logic before deploying to production
Automating regression testing for complex Salesforce AI agent configurations
Identifying and fixing gaps in topic descriptions to improve agent intent recognition

About

Key Features

Comprehensive 100-point scoring system across five critical validation categories
YAML-based test spec generation for standardized and repeatable agent testing
32 GitHub stars
Interactive simulated and live preview modes for debugging agent behavior
Automated test execution with real-time coverage metrics for topics and actions
Agentic Fix Loop for autonomous resolution of failing test cases

Use Cases

Verifying agent routing accuracy and action logic before deploying to production
Automating regression testing for complex Salesforce AI agent configurations
Identifying and fixing gaps in topic descriptions to improve agent intent recognition