Agent TodoWrite Inconsistency: Implement Strict vs Interpretive Mode

Bug Description
Agents are losing track of their work constantly despite a TodoWrite and strict instructions. How the agent writes thing in the TodoWrite seems to be "freestyled" unless told to be precise (which I thought I was doing already in the scripts). I believe that there needs to be a little bit more consistency in how the agents are instructed to use the to-do right tool. At least let us implement something at a configuration level where we can say strict vs interpretive. I find that the agent frequently just reorders tools, displays them, and displays to-dos in weird orders, and is generally an inconsistent tool to try and get the agent to use.

Environment Info

• Platform: linux
• Terminal: vscode
• Version: 1.0.83
• Feedback ID: 8b288d72-5797-4619-b6f9-ddfe2c3d4706

Errors

[]