review-gatelisted

# Review Gate **This is exloom's one enforced mechanism — the rest of exloom is discipline a model can skip.** Every other skill is guidance the model reads and *chooses* to follow, and sometimes it won't. This gate is different: when it is turned on, the `Stop` and `PreToolUse` hooks are run by the Claude Code harness — not by the model. The `PreToolUse` hook *physically blocks* a `git push` / PR (shell **and** GitHub MCP), and the `Stop` hook blocks a completion claim it recognizes, until the tier's evidence is in `.claude/reviews/<branch>.md`. That is the difference between *hoping* review happened and *guaranteeing* it did, and it is the part of exloom you cannot reproduce by prompting. (Enforcement is text-based command/tool matching plus the checklist's git checks: it catches the common `git push` / `gh pr create` forms and the listed GitHub MCP tools; a deliberately obfuscated shell command or a raw API call can still evade it, and — because matching is textual and fail-closed — a benign command that literally contains the words `git push` / `gh pr create` (a commit message or echo about pushing) can occasionally be over-blocked during the review window (bypass with `EXLOOM_REVIEW_SKIP=1`, or rephrase). It's a cooperating-team gate with a documented `EXLOOM_REVIEW_SKIP` bypass, not an adversarial security boundary.) ## Why this exists Consider a large, multi-batch refactor that passed every review it was given — per-batch code-quality reviews, per-batch spec-complia