Hard Lock Out Mechanics

I have run into issues where the chat agent that is engaged for editing a storybible or outline element targets instead changing chapters without a clear UI/UX indication of what was being edited or why.

In the specific case where this was the biggest problem, the agent had been told to resolve a contradiction in its characterization of a storybible/outline element, and instead of fixing that, it rewrote the prose in the chapter that was correct to conform to its incorrect characterization. The changes didn’t appear in the obvious places, so I had to go through the chapters to locate where the changes had been made. The changes were marked in purple, but this does not allow for differentiation of the those changes from other purple text that is in process from previous unrelated edits.

When I set an agent to working on the outline of a single chapter, the agent should not be able to abitrarily decide to start editing the manuscript of unrelated chapters or established storybible elements.

I can see the value in that being a feature that can be enabled, but in general, the agents need to be more focused because any non-trivial writing project will be too much context for them to meaningfully use without dilution or significant operational confusions.

I am aware of the “sudowrite can’t see this” toggles on some of the storybible elements, but this is more basic than that. I should be able to fix the exact context that the agent is allowed to edit, and the agent should not have the ability to override that. Certainly not silently.