What problem does it solve? Coordinating multi-perspective code reviews manually is slow and error-prone. This Skill automates the fan-out of focused code reviews by resolving a code artifact to a unified diff, dispatching one review-code subagent per requested focus in parallel, and returning a consolidated summary of which review files were written and which reviewers errored. ## Core Features & Use Cases - Artifact-to-diff resolution: Accepts a .diff/.patch file, a single source file, or a directory, and derives a unified diff via git diff against the worktree base branch, with an empty-diff guard that refuses to dispatch when there is nothing to review. - Parallel focus fan-out: Dispatches one review-code subagent per focus (architecture, accuracy, dx, security, performance, etc.) in a single batch, each writing an rNN-{focus}.md review file with automatically derived sequential numbering. - Error surfacing summary: Produces a status table showing which review files landed and which reviewers errored, with an overall ok/failed headline so calling skills can decide whether to proceed. - Use Case: During plan execution, execute-plan calls this Skill with a baseline focus list of accuracy, architecture, and dx to get three independent review reports on the just-written code before merging. ## Quick Start Ask the agent to validate the code changes in the current worktree with the focuses accuracy, architecture, and dx, writing the review files into the reviews directory.