grok-adversarial-reviewlisted

# Adversarial review via `grok` (Grok Build) The value of this leg is a **sixth accounting family**: Grok (xAI) is none of GPT/Gemini/Claude/DeepSeek/Nemotron, so it can satisfy the cross-family rule when the other paid families are already spent on authorship. It is a paid weekly pool — schedule it like the codex leg's quota, never like the free pool's best-effort capacity. ## Before the first run of a session Read **[`references/grok-runtime.md`](references/grok-runtime.md)** (same directory). It holds the family-level mechanics — the always-approve config trap, the three read-only layers and which one is actually load-bearing, the old-kernel sandbox no-op, the prompt-file rule, the measured quota fingerprint — plus the UNVERIFIED list this young family still carries. **This family has zero calibration-journal rows**: until the first verified hit rates land, treat its verdicts as one more pair of eyes, not the gate leg, and record every run. ## Establish an immutable review target **One frozen directory per reviewer, at the exact commit, that nothing else touches — no lead activity inside it.** Not "whenever possible": a reviewer reads the WORKING TREE, not your commit. Measured — a round ran mutation testing in the same worktree mid-review and the reviewer opened a CRITICAL on a mutated, non-compiling file it was never meant to see. Freeze it with the suite's tested helpers instead of hand-rolling the shell; every bug ever found in this step was in a hand-rolled copy