opencode-adversarial-reviewlisted

# Adversarial review via `opencode` (free pool) The value here is an **additional model family at zero quota cost**: DeepSeek and Nemotron are neither GPT, Gemini, nor Claude, so this is the cheapest way to satisfy the cross-family rule's second reviewer on HIGH-risk work. The trade-off is best-effort capacity — congestion is normal, so this leg is the *additional* opinion or the unhurried gate, never the time-critical one. ## Before the first run of a session Read **[`references/opencode-runtime.md`](references/opencode-runtime.md)** (same directory). It holds the family mechanics — the free-model catalogue and the stealth-model family-unknown caveat, the two silent permission traps (zero-commit project binding, last-match-wins ordering), congestion behavior, and the audit log lines. This file assumes them and covers only the review role. ## 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. The suite-root resolver and both calls — `freeze-target.sh` to create it, `verify-target.sh` before AND