Here is a brief, actionable plan for testing long-context prompt chunking. This plan focuses on **functional correctness** (does the model still get the right answer?) and **context integrity** (does the model lose critical information?).

### Phase 1: Baseline & Data Preparation
**Goal:** Establish a control