What problem does it solve? Single-model code reviews miss defects due to each model's blind spots and produce false positives that waste reviewer time. This Skill runs two independent AI reviewers on the same diff and verifies every finding against the actual code before reporting it. ## Core Features & Use Cases - Parallel multi-model review: Runs a Claude review subagent and OpenAI Codex CLI concurrently on the same diff using an identical reviewer brief, so findings are directly mergeable. - Orchestrator verification and adjudication: Every claimed finding is checked against the reviewed revision, deduplicated by root cause, tagged by which model found it, and ranked by severity. - Flexible review targets: Reviews a GitHub PR by number or URL, staged changes only, or the current branch against its merge-base, with an optional --fix step that applies verified fixes. - Use Case: Before merging a pull request, ask for a multi-model review to get one de-duplicated report showing which defects both models agreed on, which each caught uniquely, and which findings were dropped as false positives. ## Quick Start Ask the agent to run a multi-model review of the current branch, or pass a PR number such as "review-multi-model 123" to review that pull request.
