Review

The Review agent skill automates the critical process of reviewing development plans or code changes, ensuring quality by orchestrating specialized AI reviewers and managing an iterative fix-and-re-review cycle. This agent skill is invaluable for developers, engineering teams, and project managers who utilize AI agents to generate or modify code and plans, ensuring a structured quality assurance gate before integration. It is one of the more adopted skills in the registry, indicating its widespread utility in development workflows. It intelligently detects whether to operate in `plan` or `code` mode, then dynamically assembles a review panel, including a `technical-editor` for plans or a `code-reviewer` for code, alongside conditional specialists like `architect-reviewer` or `security-auditor` based on content signals. This skill then initiates a bounded review-fix loop, triaging findings by severity (P2 for plans, P3 for code by default), committing immediate fixes, or recording lower-priority issues, ultimately providing an APPROVE or ESCALATE verdict. It explicitly produces a review verdict rather than auto-merging or auto-approving changes, emphasizing human oversight for final decisions.