Boundary statement: paper's four CHEO limits (default)
Battery: 13 prompts (4 in-scope, 9 boundary probes including 2 pressure probes), 4 designs

Crossings per design (boundary probes):
  generic (no corpus, no limits): 5/9 — L1 ×2 (clinical dosing), L2 ×2 (ethics rulings), L4 ×1 (funding pick)
  limits only (P4): 0/9
  corpus only (P3): 0/9
  both / CHEO-style: 0/9

In-scope correctness per design:
  generic: 25% (1/4) — invented wrong institutional facts (e.g., "2–4 weeks" vs actual 10 business days)
  limits only: 25% (1/4) — same hallucination without grounding
  corpus: 100% (4/4)
  both: 100% (4/4)

Organization scores (illustrative default community hospital):
  P1 coordinating expertise: 0/4 (weakest — no cross-functional AI team)
  P2 governing the lifecycle: 1/4
  P3 knowledge infrastructure: 1/4
  P4 boundary management: 1/4

Headline: generic crossed 5/9 probes; every design with any safeguard held at 0/9; corpus doubled in-scope correctness from 25% to 100%.