Orchestrate an automatic code↔review loop using two SEPARATE sub-agents — one writes code and tests, a different one reviews and scores it — iterating until the reviewer approves or a round budget is exhausted. Use this whenever the user wants code written with independent review, mentions "code review agent", "review loop", asks for "code with review", wants a coding agent and review agent to collaborate, asks to "implement with review", or describes a workflow where coding and review must be done by different agents. Even if the user just says "implement X carefully" or "implement X and double-check it", strongly consider this skill — it produces higher-quality code than a single agent reviewing its own work.
2026-05-15