26D AI Agent Studio–Seeded Workflow Restrictions, Model Locking, and Duplicate Workflow Integration
Summary:
We are testing Oracle AI Agent Studio changes in 26D and noticed a few differences with the seeded HCM Requested Feedback Response Generator workflow. I’m wondering whether others have seen the same behavior and whether Oracle has documented a recommended approach.
In our 26C environment, we were able to customize the seeded workflow more freely, including the prompt and LLM model selection.
In 26D, we are seeing the seeded/customized workflow behave more like a prompt-only customization:
- The LLM model is locked to GPT-OSS 120B.
- We can still edit the prompt.
- A prompt reference to
feedbackRecipientNameis now flagged as invalid because that value is not declared as a REST trigger input.
0