agent-deck

Run iterative heterogeneous review rounds for complex code, plans, or mixed changes with two confirmed reviewer slots. Use for deep race, lifecycle, architecture, security, performance, plan-gate, or design-to-code risk, including 深度 review, 双对抗 review, review agent 深挖, 再 review 一轮, 深挖整体改动是否符合预期, and plan 评审. Continue routine in-scope review/fix rounds autonomously; ask the user to review only when a proposed remedy requires an architecture-level or similarly major decision.

Run exactly one independent review round followed by one bounded rebuttal round with two confirmed heterogeneous reviewer slots. Use for focused code, plan, prompt-asset, technical-decision, agent-validation, or overall-change checks, including 简单 review, 轻量 review, 帮我 review, 这个对不对, 对抗一下, 决策评审, and 整体改动是否符合预期. Present the evidence and recommendation to the user for final judgment; do not start a fix-and-re-review loop.

Agent Deck built-in skill self-check. Use when the user calls /agent-deck:hello-from-deck or asks whether Agent Deck, the deck plugin, or the Agent Deck skill chain is loaded, including Chinese prompts such as deck 自检. Replies with a fixed readiness message, cwd, and timestamp.