What problem does it solve? Reviewing local changes, branches, or entire codebases before committing is time-consuming and easy to do inconsistently. This Skill orchestrates a structured, read-only review of code as it sits on disk, applying specialized review lenses and returning a single severity-grouped report without editing files, running tests, or publishing anything. ## Core Features & Use Cases - Scope-parameterized review: Resolves the review target from your request — working diff, branch diff against a base, unpushed commits, a specific path, or a whole-codebase audit with chunking. - Specialized review lenses: Selects lenses (correctness, idiom, architecture, performance, simplicity, comment, sql) based on what the change actually touches, with support for C, Go, Rust, and Zig plus SQL files. - Delegated or direct execution: Dispatches one read-only reviewer sub-agent per lens in parallel for large scopes, or runs a single direct pass for small ones, then synthesizes and deduplicates findings by root cause. - Use Case: Before pushing a feature branch, ask for a review of your changes against main; the Skill detects the languages, selects the relevant lenses, and returns one report grouping findings by High, Medium, and Low severity with proposed fixes. ## Quick Start Ask the AI to review my local changes against the main branch and report any correctness, performance, or idiom issues without modifying any files.
