Hello Wrike Community!
I am trying to build an AI Agent to enforce a naming convention, and I keep hitting a wall with how the AI handles bulk actions on subitems.
My Goal: When a Project's status changes to "In Progress," I want the AI Agent to append the Project's name to the end of all Task names within that project (if it isn't there already).
My Agent Setup:
- Location: Built at the Folder level.
- Trigger: Status changes to "In Progress" (Trigger location: The subitems of the folder, meaning the Projects).
- Action: Change item name (Action location: The subitems of the triggered project, meaning the Tasks).
- Instructions: "Identify the specific project that just triggered this status change. For each task inside this project, check the task name. If the task name does not end with ' - [Project Name]', append it..."
The Issue: The agent consistently fails. When I check the reasoning log, the AI states: "The data does not provide the current names of these tasks... it is not possible to determine which tasks need renaming."
It appears that when an AI Agent is triggered by a parent item (the Project) and asked to iterate over subitems (the Tasks), Wrike's backend does not actually pass the metadata (like the current names of those subtasks) into the AI's context window. The AI knows the tasks exist, but it is flying blind regarding their properties.
My Questions:
- Is this a known limitation of the current AI Agent context window?
- Is there a way to force the Agent to "read" subitem data during a parent-triggered bulk action?
- Are there plans to expand the context passed to the LLM so bulk updates on subitems become viable?
(Note: I know I can bypass this by triggering the agent on individual "Task Creation" instead of the Project status change, but I am specifically trying to solve this for bulk updates upon project launch).
Thank you!
