claude-review

# claude-review — independent second opinion from Claude Self-review shares the author's blind spots: the reviewer is the same model that wrote the code, working from the same assumptions. A different-vendor model (Anthropic Claude) has a different training distribution, so its blind spots do not overlap with Codex's. One side writes, the other side challenges — a cheap QA pass. This skill runs the Claude Code CLI (`claude -p`) as a **read-only** reviewer and presents its output **verbatim**. Claude is given only read tools, so it cannot edit your code. ## When to use - Before merging an important diff (last gate). - After writing a spec / tests — ask Claude to find ambiguity and weak assertions. - A hard bug whose root cause is unclear (independent diagnosis). ## Step 0 — Preflight (binary) ```bash command -v claude >/dev/null 2>&1 || { echo "[claude-review] Claude Code CLI not found. Install from https://claude.com/claude-code (then sign in). Skipping." exit 0 } ``` If this prints the skip message, tell the user Claude Code is not installed and stop. ## Step 1 — Resolve review scope and capture the diff ```bash ROOT=$(git rev-parse --show-toplevel 2>/dev/null) || { echo "[claude-review] not in a git repo"; exit 0; } cd "$ROOT" BASE=$(git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's|refs/remotes/||') if [ -z "$BASE" ]; then if git rev-parse --verify -q origin/main >/dev/null 2>&1; then BASE=origin/main elif git rev-parse --verify -q origin/mast...