Diagnose Agentforce Agent Errors

1. Checks a reported failure against the agent's live setup. In one case it confirmed that topics, actions and planner were sound and traced a persistent "Something went wrong" error to an inactive bot version.
2. Traces why the agent's logic misbehaves. One example is a counter that did not count failed order verifications.
3. Writes an implementation plan for the fix.
4. Applies the fix in a new agent version through the builder's AI assistant, then saves and commits it. Activation is left to the user.
5. Diagnoses runtime failures, such as an agent service that never invoked the model.
6. Hands activation to the user in Setup when tool permissions block it from the command line.

Related: Diagnose Agentforce Action Failures and Diagnose Agentforce Permission Errors.