1实现原理 · 为什么它能做到
定位为纯 explore 阶段:不承诺结构、不建大纲,跑一场『产碎片』的 grilling 会话,把探索空间撑宽。
skills/in-progress/writing-fragments/SKILL.md
This is pure **explore**: widen the space of what could be written without committing to structure. Committing is _exploit_, a separate skill's job. Run a grilling session that produces fragments, interviewing the user relentlessly about whatever they want to write about. Imposing phases, outlines, or article structure is out of scope here.
explore/exploit 分界是仓库 writing 系列的组织原则:碎片阶段禁止一切结构承诺。
捕获纪律:连用户最初那句 prompt 也要当首批碎片收录;碎片追加到单个 markdown 文件,首次写入只放一个 H1 工作标题,无 metadata/TOC/日期。
skills/in-progress/writing-fragments/SKILL.md
Capture fragments from the very first thing the user says, including the initial prompt. On first write, put a single H1 at the top with a working title (it can change later) and nothing else: no metadata, no TOC, no date.
格式即纪律:'Fragments are separated by a horizontal rule (`\n---\n`). No headings inside the body. No tags. No order beyond the order they were added.'
碎片质量标尺 = 作者可读即可,不需要冷读者能懂:'is this a piece of good writing?' 而非 'is this a self-contained argument?'。
skills/in-progress/writing-fragments/SKILL.md
It must be _readable by the author_ (the author can tell what it means), but it does not need to define its terms or be comprehensible to a cold reader. The bar is "is this a piece of good writing?", not "is this a self-contained argument?"
允许半成品(half-thought)、vignette、引用、妙句等异质碎片;碎片刻意 heterogeneous。
leading word 被标为最有价值的碎片类型:当对话围绕重复出现的念头,要主动铸一个命名该念头的词/隐喻,为后续 exploit 的结构、转场与标题奠基。
skills/in-progress/writing-fragments/SKILL.md
Of these, the leading word is the most valuable fragment to land. It is load-bearing: name the right one in explore and it shapes the structure, the transitions, and the title later, paying dividends through the entire exploit phase. When the conversation circles a recurring idea, push to coin a word for it.
例证用 _tracer bullets_ / _fog of war_ 这类命名整个模式的词——作者的方法论偏好。
节奏纪律:静默追加、不逐条请求许可;每次写前重读磁盘、绝不整体覆盖,只追加或按用户要求就地改某条。
skills/in-progress/writing-fragments/SKILL.md
Append silently. Don't ask permission for each fragment. Mention what you added in passing ("adding that"), but don't interrupt the conversation with save dialogs. Before every write: re-read the file from disk. The user may have edited, reordered, or deleted fragments between turns, so preserve their changes. Never overwrite the file; only append (or, if the user asks, edit a specific fragment in place).
对话流畅性优先:写入动作降到最低打扰;文件状态以磁盘为准。
2核心能力
01以 grilling 会话挖掘用户(含首次 prompt)的写作碎片
02铸造 leading word:给反复出现的念头命名一个承载全篇的隐喻/新词
03单文件碎片库维护:H1 标题 + --- 分隔、无正文标题/标签/元数据
04碎片级编辑命令:cut the last one / rewrite sharper / merge those two
05跨会话记忆目标路径:未传路径时问一次并记住
4管控建议与判定依据
建议配套管控措施
面向 Agent
限定工作目录
面向管理者
保留操作记录
判定依据(文件路径与关键语句摘要)
- 本地工作目录: 文件读写与工件输出: 用户指定(或问一次后记住)的单个 markdown 文件:只追加、或在用户要求时就地编辑/删除特定 fragment
审计说明: 。仅本地文件操作:向用户指定的单文件追加碎片(明确 'Never overwrite the file; only append');增删改只发生在用户明确要求的碎片上('cut the last one' 等为一等指令)。无脚本、无网络外发、无凭证读取。注意点:写路径需用户指定;内容为模型从对话生成的文字,用户应知悉文件会被持续追加。
5第二遍独立确认
- [ok] grilling 会话引擎声明 — 'Run a grilling session that produces fragments' 原文在;调用 model-invoked 原语方向合规。
- [ok] 首次 prompt 也捕获 — 'Capture fragments from the very first thing the user says, including the initial prompt.' 原文在。
- [ok] H1 + --- 文件格式纪律 — 'put a single H1 at the top with a working title…nothing else'、'Fragments are separated by a horizontal rule' 原文在。
- [ok] 只追加不覆盖 + 写前重读磁盘 — 'Never overwrite the file; only append'、're-read the file from disk' 原文在。
- [ok] leading word 铸造指令 — 'the leading word is the most valuable fragment to land…push to coin a word for it' 原文在。
- [ok] 文件写入范围 — 仅用户指定单文件;'If the user did not pass a path, ask once where to save the document' 无其它路径。
- [ok] 外部依赖为空 — token 扫描零命中;grilling 是仓库内 skill 引用。
- [ok] 元数据 + beta 状态 — GitHub API MIT/257659/2026-09-04;本地 HEAD == 3cca18b;in-progress README 列为 beta。
6结论
适合: 适合写作前的素材勘探:脑子里有模糊主题、想边聊边攒原始材料的作者;也适合把『日常 noticings』沉淀成将来可开采的碎片库。与 writing-shape/writing-beats 串联(fragments → 素材 → exploit)使用效果最好。
不适合: 不适合想直接得到成文或大纲的用户;不适合素材已齐、只需组织(应直接进 exploit);不适合希望文件结构干净可读的场合(碎片文件刻意无组织)。
Install Agent-Ready
① Verified Mirror 2026-09-09
Option A · Direct DownloadDownload writing-fragments.tar.gz
sha256:
Option B · Copy Guide for Agentca6d39f53627a5dd…agent 读 JSON 指南后会自动从本站下载安装,无需更多说明。
② Upstream GitHub
GitHub Repo ↗
Pinned to commit 3cca18b368.