What problem does it solve? Single-pass code review misses blind spots and reflects one reviewer's biases. This Skill runs the same diff through several independent LLM reviewers, then applies lead-reviewer judgment to filter noise, verify findings against the source, and produce a verdict you can act on before merging a PR. ## Core Features & Use Cases - Multi-model adversarial review: Spawns one reviewer per configured model with a shared prompt, rubric, code-quality lens, and blind-spot lenses, so consensus findings carry the highest signal. - Lead judgment synthesis: Deduplicates and verifies every finding against the actual code, then categorizes it as Act on, Consider, Noted, or Dismissed with rationale. - Pre-PR gate and external PR review: Serves as the gate before marking a PR ready, and converts findings into classified review comments (blocking/question/nit/praise) for another author's PR, posted only after explicit approval. - Use Case: Before opening a pull request, run an adversarial review of your branch diff to catch correctness bugs, structural regressions, and missing edge-case handling that a single reviewer would miss. ## Quick Start Ask the assistant to interrogate review the current branch diff before opening the pull request.
