1实现原理 · 为什么它能做到
以 grilling 为引擎的有状态封装:直接跑一个『只产 workflow 规格』的 /grilling 会话,复用其 relentless + 一轮多问 + 每题推荐答案的纪律。
skills/in-progress/loop-me/SKILL.md
Run a stateful `/grilling` session whose only output is **workflow** specs. Use the grilling discipline (relentless, a round of questions at a time, a recommended answer attached to each) aimed at the vocabulary and goal below. Create, edit, and delete specs as the grilling resolves things.
注:这里在做什么:loop-me 不发明新的提问机制,而是给 grilling 套上领域词汇表(loop/workflow/trigger/checkpoint/push right/brief)与产出约束(workflow specs),并授权在拷问过程中直接写工作区文件。
loop lens 识别可委托对象:把用户生活的重复模式(职业生涯/一周/早晨/单个重复活动)看作 loop,loop 可预测性 = 值得委托的理由,且要主动提出用户自己没注意到的 loop。
skills/in-progress/loop-me/SKILL.md
A **loop** is a recurring pattern in the user's life: their career, their week, their morning, a single repeated activity. Picturing a life as loops within loops reveals how predictable its activities really are, which is what makes them worth **delegating**. Use the lens to find loops worth specifying, and propose ones the user hasn't noticed.
注:workflow 是 loop 的规格实体:'A **workflow** is the spec of one loop, made real. You run a workflow on a loop: the loop is its running instantiation.'
词汇表克制原则:共享语言只在 workflow 需要时使用、绝非 checklist;结构上零强加(AI/checkpoint/schedule 都不是必需,除非拷问证明需要)。
skills/in-progress/loop-me/SKILL.md
A shared language, reached for only when a workflow calls for it: never a checklist. **Mandate nothing structural**: a workflow needs no AI, no checkpoint, and no schedule unless the grilling shows it does.
注:这是对『所有 workflow 都必须自动化/AI 化』的显式反对——纯人工 loop 也在范围内。
两个核心设计原则:push right(把人工 checkpoint 尽量推迟,一次问、晚点问、备齐一切)与 brief(checkpoint 呈现给用户的是决策就绪的简报而非原始输出)。
skills/in-progress/loop-me/SKILL.md
**Push right**: defer the checkpoint as far as it will go. Do maximal work before involving the human, so they are asked once, late, with everything prepared. **Brief**: what a checkpoint presents, a tight, decision-ready summary (what was produced, why, and a link down to the asset itself), never the raw output. The user reads a brief, not a draft. Speed of review is imperative.
注:这定义了被产出 workflow 的行为美学:最大化无人运行段,人工介入点少而精。
完成标准 = 实施者可无问建造:拷问一直进行到『implementer agent 能不问一个问题就照规格建出来』为止,问题未尽则未完成。
skills/in-progress/loop-me/SKILL.md
A workflow spec is done when an implementer agent could build it without asking a single question. Grill until then; nothing is done while a question remains.
注:把 grilling 的『frontier 空 = 完成』翻译到本领域:规格的未知面全部清零。
双文件工作区作为跨会话状态:workflows/*.md 每 loop 一份规格(source of truth),NOTES.md 存用户世界的原始观察与术语;NOTES 空/薄时先访谈用户的世界再开始规格化。
skills/in-progress/loop-me/SKILL.md
The workspace - `workflows/*.md`: one spec per workflow. - `NOTES.md`: raw notes on the user's world, the tools they use, the channels they process, and their own terminology for both. When it is empty or thin, interview them about their world before specifying anything. Sharpen fuzzy terms into canonical ones as they surface, and record them here.
注:in-progress README 称其为 'over multiple sessions, using the current directory as a stateful workspace'——状态 = 这两个文件。
2核心能力
01识别并提议用户未察觉的可委托 loop(生活重复模式 → workflow 候选)
02多会话 stateful 拷问:跨会话维护 workflows/*.md 与 NOTES.md
03按 loop 词汇表产出规格(trigger/checkpoint/push right/brief),结构零强加
04世界建模前置:NOTES 空时先访谈用户的世界/工具/术语再规格化
05以『实施者可无问建造』为完成门禁
4风险提醒 风险提醒:黄色 · 留意使用
风险提醒:黄色 · 留意使用
- beta、可随时消失 — in-progress bucket:不入 plugin、无 docs、'they can change or disappear without warning';依赖它的流程需自行备份。
- 授权增删改工作区文件 — skill 明文授权在拷问过程中 create/edit/delete workflows/*.md——delete 属破坏性本地操作(范围限定于该管理目录,非任意路径/清盘),故分档为黄色;若跑错目录或用户误触发,会改动非预期文件。应在正确工作区启动。
- 产出即真相的隐患 — workflow 规格标称 source of truth;模型访谈 + 书写可能引入对用户生活的错误假设,用户不审阅就实施会把坏规格固化。
- 方法论强个人色彩 — loop lens 是作者个人对『生活可委托性』的框架;不认同该视角的用户可能觉得访谈方向跑偏。
风险提醒:黄色,留意使用。行为 = 拷问用户 + 在当前工作区增删改 workflows/*.md 与 NOTES.md:skill 明文授权 'Create, edit, and delete specs as the grilling resolves things',delete 属破坏性本地操作(范围限定于其管理的 workflows/*.md,非任意路径/清盘),按统一分档取高一档为黄色。无网络外发、无凭证读取、无脚本执行。使用建议:在专用工作目录启动、先确认目录正确,并审阅规格后再交实施。
5第二遍独立确认
- [ok] 复用 /grilling 引擎声明 — SKILL.md 首句 'Run a stateful `/grilling` session whose only output is **workflow** specs' 存在;grilling 为 model-invoked 原语,调用方向合规。
- [ok] 文件操作范围(workflows/*.md、NOTES.md create/edit/delete) — 原文 'Create, edit, and delete specs as the grilling resolves things.' + The workspace 段;无任何工作区外路径/绝对路径/删除指令。
- [ok] user-invoked 双清单 — disable-model-invocation: true + policy.allow_implicit_invocation: false 均在;argument-hint 亦在 frontmatter。
- [ok] loop lens / 词汇表 / push right / brief 概念齐全 — 逐段存在(A **loop** is…、Push right、Brief、Trigger、Checkpoint)。
- [ok] 完成门禁(implementer 无问可建) — 原文 'A workflow spec is done when an implementer agent could build it without asking a single question.' 存在。
- [ok] 外部依赖为空 — token 扫描零命中;无 CLI/API/网络/包引用。grilling 是仓库内 skill 引用非外部依赖。
- [ok] 元数据 + commit — GitHub API MIT/257659 stars/2026-09-04T08:45:43Z;本地 HEAD == 3cca18b。
- [ok] beta 状态标注 — in-progress/README.md 将 loop-me 列于 in-progress 并声明 excluded from plugin、'can change or disappear without warning',与 verdict 风险描述一致。
6结论
适合:适合想把自己重复的工作/生活活动系统化委托给 agent、且愿意接受多轮拷问访谈的用户;尤其适合在专用工作目录里跨会话累积『个人 workflow 库』的场景,产出可直接交给 implementer agent 落地。
不适合:不适合单次快速问答(它是有状态、要写文件的长会话);不适合不想让 skill 动工作区文件的人;不适合团队共享仓库(规格写作带强个人视角,且 beta 不稳定);期望产出通用工程文档(spec/ADR)的人应改用 to-spec/grill-with-docs。
安装 agent 直装可复制
① 本站镜像 更新 2026-09-09
方式 A · 人下载镜像包下载 loop-me.tar.gz
sha256:
方式 B · JSON 格式安装指南,复制给 agent7815f818e390fcec…agent 读 JSON 指南后会自动从本站下载安装,无需更多说明。
② 上游 GitHub · 原始来源
来源信息 GitHub 原始
作者 / 仓库mattpocock / mattpocock/skills
原始 GitHubmattpocock/skills ↗
Stars257659
最近推送2026-09-04
本 skill commit
3cca18b368许可MIT(GitHub API spdx MIT;仓库根 LICENSE: Copyright (c) 2026 Matt Pocock)
本站信息
收录日期2026-09-06
分类基础工具与工作流
侦查报告AI 侦查 · 2 遍 · 2026-09-06
本站镜像与 GitHub 原始是不同来源:本站锁定 commit 快照经 /r2 分发;GitHub 为实时上游,内容可能已更新。