https://dextermayhewjd.github.io/projects/learn-claude-code/s07/Recent content in S07 Skills：从最小示例学习技能编写 on dextermayhewjdHugozh-cnhttps://dextermayhewjd.github.io/projects/learn-claude-code/s07/01-structure/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/01-structure/采用官网 Quickstart 的 roll-dice 示例，通过一次掷骰子认识 name、description 和正文指令。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/02-real-expertise/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/02-real-expertise/从真实任务和项目资料中提炼步骤、纠正、输入输出与约定，理解 SOP 如何成为可复用的 Skill。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/02-discovery/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/02-discovery/看清个人、项目、嵌套目录和插件 Skill 的发现方式，以及名称冲突的处理。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/03-content-selection/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/03-content-selection/沿用 roll-dice，用两份完整 Markdown 对照，区分给读者的解释、Agent 已知的常识和任务真正需要的指令。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/03-invocation/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/03-invocation/区分发现、菜单可见、模型可见和真正调用，解释两种调用控制字段。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/04-task-scope/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/04-task-scope/沿用学习小节编写示例，从完成结果出发，识别范围过宽与过窄，再把任务边界写回完整的 Skill。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/04-context/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/04-context/跟踪描述、正文、参考文件的可见时间，解释加载后的保留、重复调用和压缩。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/05-control-calibration/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/05-control-calibration/按操作的脆弱程度调整指令的具体程度，给出默认方案，并描述同类任务可复用的处理流程。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/05-rendering/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/05-rendering/用实际调用说明参数替换、技能目录变量和动态命令注入，区分预处理与模型行动。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/06-permissions/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/06-permissions/分清 allowed-tools、disallowed-tools 与权限策略，说明 Skill Hooks 的执行和持续时间。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/07-references/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/07-references/从官方多文件例子出发，区分主文件与配套资料，写清相对链接、读取条件，并检查按需读取是否发生。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/07-subagents/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/07-subagents/比较 inline、fork、后台和等待模式，并区分上下文隔离与工作目录隔离。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/08-evaluation/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/08-evaluation/用明确的输入和结果标准分别验证自动触发与任务质量，定位发现、渲染和执行阶段的问题。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/09-agent-loop/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/09-agent-loop/学完独立机制后，回顾 S06，再用总图和伪代码说明目录、调用与正文交付如何进入原循环。https://dextermayhewjd.github.io/projects/learn-claude-code/s07/10-appendix/Mon, 01 Jan 0001 00:00:00 +0000https://dextermayhewjd.github.io/projects/learn-claude-code/s07/10-appendix/用可下载的 review-diff 练习工程覆盖 S07.1–S07.9；每份文件标明位置、讲解章节和注意点。