Done. Here's the summary:

**Experiment**: Brody parameter flow along the prime sequence (72 windows of 5000 gaps, up to p = 2M, with shuffle + Cramer controls).

**Key finding**: Primes flow toward Poisson, not GUE. beta(p) = 0.64 - 0.030 * ln(p), R^2 = 0.78, z = -2.42 vs shuffle.

**Two-channel decomposition of the flow**:
- 82% magnitude (gap distribution becomes more exponential as primes thin out)
- 18% ordering (sequential correlations add a small persistent depression)

**The structural insight**: This flow is **perpendicular** to the spectral rigidity finding (beta increases with scale L). The GUE/Poisson boundary is not a point — it's a 2D trajectory in (N, L) space with opposing gradients.

**Prediction**: beta = 0 at p ~ 1.3 × 10^9 (testable by extending the sieve).

**Files**:
- Script: `tools/exp_brody_flow.py` (reusable with --n-max, --window, --step, --n-shuffle)
- Data: `tools/data/brody_flow.json`
- Report: `tools/data/reports/agent_20260429_1013.md`
- Seme updated with BRODY_FLOW tension and META consecutio refined.