Orchestrator
Let AI research you back—answer ~15 multiple-choice questions to get YOUR AI config (CLAUDE.md / AGENTS.md / custom skills), instead of copy-pasting templates. By FantasyMax.
npx -y skills add Fantasymax/easy-build-skills --skill orchestratorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Orchestrate a {N}-step workflow that {B1 一句话定义}. This is the entry point — when the user says any of {B7 ① 顶层触发原话 1–2 句}, this skill walks them through {sub-skills 列表} in order. Trigger especially when the user mentions {B7 整体触发原话} or wants to "do the whole flow". Do NOT use for: single-step requests (use the corresponding sub-skill directly).
SKILL.md
3.3 KB, as published. Nobody here has run it
{Skill 链标题}
这是 {用户名} 的 {B1 一句话定义} 的 orchestrator。本 skill 不直接做任务,而是按次序驱动 N 个 sub-skills。 生成日期 {YYYY-MM-DD} | V1(快速版)
Overview
{B1 一句话定义原话}
整条链分为 N 步:
/{chain-name}-step-1-{name}— {第 1 步做什么,一句}/{chain-name}-step-2-{name}— {第 2 步做什么}- ...
每一步独立可跑(用户可以中间插入 / 跳过 / 重跑),也可以一气走完。
Quick Reference
| 步骤 | sub-skill | 输入 | 输出 |
|---|---|---|---|
| 1 | /{chain-name}-step-1-{name} | {B2 输入材料 - 第 1 步} | {第 1 步输出} |
| 2 | /{chain-name}-step-2-{name} | {第 1 步的输出} | {第 2 步输出} |
| N | ... | ... | ... |
When to use this orchestrator
- 用户想从 0 走完整条链:
/{chain-name}-orchestrator - 用户问 "{B7 整体触发原话}" 时 Claude 自动调用本 skill
不该触发:
- 用户只想做单步(如"只生成 hook 草稿")→ 让 Claude 直接调
/{chain-name}-step-N-{name} - 用户想做与本链无关的 generic 任务
Workflow(链次序)
[Phase 1:链入口]
1. 接收用户输入({B2 输入材料 - 起点})
2. 简短复述:"我会按 N 步处理:[1] [2] [3],每步走完会给你看,你可以选'继续'或'停在这里改'"
[Phase 2:调度执行]
3. 调用 /{chain-name}-step-1-{name},传入用户原始输入
4. 拿到 step-1 输出 → 展示给用户 → 等"继续"或修改指示
5. 调用 /{chain-name}-step-2-{name},传入 step-1 输出
6. 拿到 step-2 输出 → 展示 → 等"继续"
7. ...直到最后一步
[Phase 3:收尾]
8. 整链跑完后总结:"链已跑完。你可以:① 接受全部 ② 回到第 N 步重做 ③ 把整条产出导出"
Hard Rules
- MUST 在每步之间暂停等用户确认(B11 选 ② 时通常意味着用户希望见每步结果,不能一气端)
- MUST 在出错时告知"在第 N 步失败了,原因是 X,可以从第 N 步重启",不要静默重启整链
- NEVER 跳过某步(如果某步对用户太简单,Claude 会觉得"我直接用通用能力做就行"——那就该把那步从链里删掉,而不是动态跳过)
- ALWAYS 记录每步的输入输出(便于后续 dogfood 时找哪步出了问题)
When to ask
- 第 1 步开始前:「你这次是想跑全链还是只跑前几步?」
- 每步结束后:「这步 OK 吗?继续下一步 / 还是改这步?」
- 某步用户原话和该步触发原话不匹配:「你这次的输入似乎更适合跳到第 N 步,要不要直接从 N 开始?」
Reference files
按需读:
examples/full-chain-walkthrough.md— 一次完整跑链的实例(用 B8 好范例对应场景)references/glossary.md— 整链共享的术语表(B6 ①)
Notes from author
{B10 ① 收尾原话;如果用户提供了链整体上的重要补充}