Add Sandbox Policy section per user mandate

AGENTS.md

@@ -139,6 +139,21 @@ env = ActionMasker(Monitor(base_env), ...)  # DON'T DO THIS
 
 ---
 
+## Sandbox Policy (User Mandate)
+
+> **From this point forward, the user has mandated:**
+
+1. **Start `cpu-basic` sandbox** at the beginning of every session
+2. **Use `cpu-basic` for**: context, writing code, writing docs, editing files, planning
+3. **Only switch to GPU sandbox** (`t4-small` or `a10g-small`) when performing **smoke tests** for training scripts
+4. **Stop GPU sandbox IMMEDIATELY** after the smoke test completes
+5. **Training tasks ONLY as HF Jobs** — never leave a training process running in a sandbox
+6. **Never leave a GPU sandbox running idle** — this wastes money
+
+**Why this matters**: A GPU sandbox at $1/hr running empty for 3 hours = $3 wasted for nothing. An HF Job at the same $1/hr actually trains for every billed minute.
+
+---
+
 ## Repo File Guide
 
 | File | What It Is |