Here's a brief plan for testing long-context prompt chunking.

## Objective
Verify that chunking a long input preserves answer quality, stays within context limits