Token导航 LogoToken导航TokenDH.com
待分类只读github未标认证来源可访问许可证需确认审计通过

structure-first-docs构建第一个文档

Agent Skill

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

总安装

349

周安装

15

GitHub Stars

27

下载量

122
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:structure-first-docs(构建第一个文档)
来源仓库:https://github.com/perhapsspy/structure-first
仓库路径:skills/structure-first-docs
安装命令:
npx skills add https://github.com/perhapsspy/structure-first --skill structure-first-docs
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/perhapsspy/structure-first --skill structure-first-docs

简介

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写,适合提炼结构、补齐章节、统一术语。

  • 它帮助把零散材料整理成可读文档,使用时需保留项目已有事实和路径。
  • 通过 npx skills add 命令从 GitHub 仓库安装,具体用法需结合 README 进一步确认。
  • 涉及对外文案时应控制语气,避免过度营销或夸大能力。
  • structure-first-docs 属于待分类类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Skill: Structure First Docs

Purpose

Restructure and review engineering documents so readers can follow the content in one top-down pass. Preserve source meaning first. Do not invent facts. Do not finalize decisions that were not finalized in the source. Preserve useful source form too. Do not flatten emails, checklists, ADRs, or handoff notes into one generic output shape unless readability clearly improves.

Modes

  1. Compose
  • Draft/rewrite documents in readable top-down flow.
  1. Review
  • Review documents and return issues/risks first.

When to Use

  • Design docs, refactor/migration plans, runbooks, PR narratives
  • Technical doc review before implementation/share

Do Not Use

  • Marketing/brand copy
  • Strict compliance/legal content owned by policy/legal teams

Core Rules

  1. Fix intent in one sentence first.
  2. Build Primary Reader Flow from existing source content.
  3. Source fidelity first: every statement must be traceable to source text.
  4. Do not add new facts, constraints, metrics, owners, dates, or policy values unless explicitly provided.
  5. Do not upgrade tentative language into confirmed decisions.
  6. Preserve tentative wording strength. might, seems, looks like, likely, maybe, and similar language should stay at similar strength.
  7. If chronology is part of the meaning (timestamps, incident sequence, handoff order), preserve that chronology unless reordering clearly improves readability without losing meaning.
  8. If information is missing, keep it as Open Questions or Unknown, not assumptions.
  9. Preserve source form when it is already useful. Emails should still feel like emails; checklists should still feel like checklists.
  10. Prefer exact no-op when readability gain is marginal. Do not edit just to show structure.
  11. In review mode, always start with issues/risks.
  12. Review should clarify document state/scope/risk. Do not redesign the underlying system unless the source already frames that redesign.

Compose Structure (Adaptive)

Do not force a fixed template. Prefer source-native form first.

Use minimum sections first:

  • Intent
  • Current Facts
  • Next Actions

Add sections only when needed:

  • Decisions (only confirmed decisions from source)
  • Risks
  • Open Questions
  • Appendix

If the source is already readable enough:

  • prefer exact no-op,
  • keep the original format,
  • or make only light-touch reordering/label cleanup.

Review Structure

  • Start with Issues/Risks (severity-ordered).
  • Then provide Evidence and Required Changes.
  • Keep Open Questions/Assumptions only if needed.
  • Add a brief summary only when requested.
  • Prefer requests for clarification, scoping, or status labeling over proposing a new design.

Execution Steps

  1. Identify mode (Compose or Review).
  2. Identify the source form (email, checklist, ADR, handoff, memo, etc.).
  3. Extract source units: facts, decisions, actions, unknowns.
  4. Detect whether chronology itself carries meaning.
  5. Reorder only as much as needed for readable top-down flow.
  6. Keep source-native form when it already works; use adaptive sections only when they help.
  7. Prefer exact no-op when the gain is negligible.
  8. Verify each statement is source-traceable.
  9. Verify tentative wording was not strengthened.
  10. Convert missing or uncertain items into explicit questions.
  11. Keep output concrete and non-speculative.
  12. If direct edit is possible, edit target files directly.
  13. If direct edit is blocked, return publish-ready Markdown with failure reason.

Final Checks

  • Can a reader understand the decision path in one top-down pass?
  • Did any new fact get added that is not in source?
  • Did any tentative idea become a confirmed decision?
  • Did tentative wording stay at similar strength?
  • If chronology mattered, was it preserved?
  • Did the output preserve useful source form instead of flattening it?
  • Are sections minimal for this document?
  • Would exact no-op or lighter touch have been better?
  • In review mode, are issues/risks first?
  • In review mode, did the response avoid drifting into redesign?

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.6%
按下载量换算43

Claude

32.41%
按下载量换算40

Cursor

19.52%
按下载量换算24

Gemini CLI

9.32%
按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills