agent-command-coachlisted

# agent-command-coach > 学会有效指挥 AI Agent。不要模糊 prompt，不浪费执行时间。 > Learn how to command AI agents effectively. No vague prompts. No wasted time. --- ## 核心规则 / Core Rule 用户负责判断方向，Agent 负责执行。每条命令都要包含 Project Context Pack、明确的 Do Not、Acceptance Gate 和 Next Action。 The user orchestrates. Agents execute. Every command needs a Project Context Pack, clear Do Not rules, an Acceptance Gate, and a Next Action. --- ## 边界 / Boundaries ### This skill does - 帮用户写出更好的 agent commands / help users write better agent commands - 适配到当前 Agent / adapt commands to the current agent - 说明何时使用 slash commands、fallback prompts、subagents、long tasks 和 review gates / explain when to use slash commands, fallback prompts, subagents, long tasks, and review gates - 保持闭环 Plan → Command → Build → Review → Fix → Launch / preserve the loop: Plan → Command → Build → Review → Fix → Launch ### This skill does not do - 不强迫使用某个 Agent 工具 / does not force a specific agent tool - 不默认推荐 subagents / does not recommend subagents by default - 不为简单任务推荐 high/xhigh reasoning / does not recommend high/xhigh reasoning for simple tasks - 不允许没有人工 review 的危险自动化 / does not allow unsafe automation without human review - 不替代 Acceptance Gates / does not replace acceptance gates --- ## 工作流 / Workflow Use the sections in this file in order. Keep the output compact, preserve MVP scope, include constraints, and end with one next action. --- ## 什么时候使用 / When to use this skill 当用户不会指挥 Claude Code / Cursor / Codex / ChatGPT，或者 Age