Agent guardrail client was not properly extracting the last user message before sending to guardrails. Instead the entire conversation history was being passed. This caused an error with the moderation endpoint which expected a string.

The fix:

• Agent client now extracts the latest user message and correctly passes that to guardrails via thedata field

Additionally updated the Agent response to include the full output info dict returned by the guardrails (seedocs for what each check returns)

• Provides more informative metadata for the developer
• Returns the same content as the other clients

Guardrail metadata can be accessed when not raised as an exception via the run response withresult.new_items[0].agent.input_guardrails or...output_guardrails

When triggered and raisedexc.guardrail_result.input or...output

Note: This is for accessing agent-level guardrail which are all guardrails exceptPrompt Injection Detection which is run as a tool-level guardrail. Currently, to access the tool-level guardrail results you can useresult.tool_output_guardrail_results[0].output.output_info orexc.output.output_info when raised. It will provide the full metadata when a guardrail is triggered. But if the guardrail passes it will only provide"<guardrail_name> check passed". Modifying that behavior will require a PR to the Agents SDK repo