Now we wait for the first instance of prompt-injection via self-eval. Now that I think about it, a LLM-guided conversation that results in a review following a

Now we wait for the first instance of prompt-injection via self-eval. Now that I think about it, a LLM-guided conversation that results in a review following a pre-defined grading scale/feedback form/whatever doesn't sound like a horrible idea. I'm not talking about hooking up MCP to pull in data or feeding it data about the employee to one-shot this. Rather the LLM asks a series of questions, with follow-ups where it makes sense, and tries to hone you responses into useful, actionable feedback.