26A AI Agent – Inconsistent input variable binding resolution: configuration gap or defect?

Summary:

When testing an AI Agent Studio Agent Team Workflow in the AI Agent Studio Lab, I am seeing inconsistent results with input variable bindings:

• Agent Node block traces show that workflow-level input variables resolve correctly, but the Agent Input variables referenced in the agent prompt do not resolve consistently.
• Using literal values in the agent prompt works reliably, confirming that the issue occurs specifically when input variable bindings are used.
• The behavior is intermittent. Sometimes the bindings are applied successfully, other times they are not.

Is this a configuration gap or a defect?

See the current agent prompt and trace screenshot below for reference.