Builder validation checks and trace previews for troubleshooting
Key Takeaway
Validation checks catch build problems when you save. To learn why the agent chose a subagent or an action during a conversation, use the Trace panel in Preview.
Agentforce Builder gives you two kinds of troubleshooting help. Validation checks catch problems in how the agent is built, before you commit it. Testing in the builder shows how the agent actually reasoned during a conversation, so you can see why it picked a subagent or ran an action.
Validation checks
When you save changes, Agentforce Builder runs validation checks in both Canvas view and Script view. Your agent's configuration is the same in both views, so a problem found in one view is a real problem in the agent.
The console lists the errors and warnings that could interfere with your agent's performance, such as syntax errors or unexpected values.
You can commit an agent that has warnings, but you can't commit one that has errors. A warning flags a risk. An error blocks the commit until you fix it.