What problem does it solve? Single-model code reviews can miss defects or reflect one model's blind spots. This Skill runs multiple independent AI reviewers over the same Git diff, verifies they inspected identical content, and merges their findings into one deduplicated, severity-ordered report. ## Core Features & Use Cases - Multi-model review orchestration: Runs Claude Code and native Codex reviewers in parallel by default, with Cursor Agent (Grok 4.6) as an opt-in third reviewer, each isolated and read-only. - Scope integrity verification: Captures a cryptographic fingerprint of the working-tree or branch diff before and after review, stopping the workflow if the scope drifts or evidence capture is incomplete. - Finding deduplication and attribution: Merges frozen reviewer reports by root cause, keeps the strongest severity, and attributes each finding to the exact reviewers that independently found it. - Optional Beads logging: With --log-to-beads, stores final actionable findings in an existing Beads issue tracker with deterministic deduplication. - Use Case: Before merging a large feature branch, ask for a comprehensive review of your working changes; receive one consolidated report listing critical-to-low findings with file locations, requirement coverage, test gaps, and per-reviewer status. ## Quick Start Ask the agent to use comprehensive-review to review your current working changes with the default Claude and Codex reviewers and report findings without editing code.
