Token导航 LogoToken导航TokenDH.com
前端设计只读github未标认证来源可访问许可证需确认审计通过

documentation-state-auditor文件国家审计员

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

360

周安装

15

GitHub Stars

公开资料未说明

下载量

120
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:documentation-state-auditor(文件国家审计员)
来源仓库:https://github.com/sjunepark/custom-skills
仓库路径:skills/documentation-state-auditor
安装命令:
npx skills add https://github.com/sjunepark/custom-skills --skill documentation-state-auditor
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/sjunepark/custom-skills --skill documentation-state-auditor

简介

审计项目文档是否清晰区分已实现功能与规划中需求的界限状态。

  • 不强制特定文档系统,追求最简结构满足项目当前阶段的表达需求。
  • 回答新读者能否快速理解现状与未来方向的关键认知问题。
  • 目标是让各类事实真相变得可见,减少猜测成本提高协作效率。
  • documentation-state-auditor 属于前端设计类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Documentation State Auditor

Use this skill when a repository's docs may blur:

  • what exists today
  • what the team intends to build
  • what is still unresolved

The goal is not to enforce one documentation system. The goal is to make those kinds of truth legible with the lightest structure that fits the project, and to make the tree readable enough that readers can move from overview to detail without guesswork.

Primary Questions

This skill should answer:

  1. Can a new reader tell what is implemented now versus what is planned?
  2. Which docs currently blur those boundaries?
  3. Is the current organization helping or hurting clarity?
  4. Does the directory tree itself reveal how the docs are organized and where a reader should start?
  5. Can a reader move from broad overview to detailed docs progressively, only loading more detail when needed?
  6. What documentation structure would fit this project with the least maintenance cost?
  7. What concrete edits would most improve clarity right now?

Core Principles

  • Distinguish kinds of truth before choosing folder structure.
  • Prefer structures that support progressive disclosure from overview to detail.
  • Prefer visible prose and headings over hidden metadata when possible.
  • Optimize for reader comprehension, not documentation purity.
  • Let the methodology vary by project maturity, team size, and documentation volume.
  • Prefer the smallest change that removes ambiguity.

Kinds Of Truth

When reviewing docs, classify content into these buckets:

  • current: factual present-state repository or runtime reality
  • target: intended or accepted future-state design
  • question: unresolved areas that should not be read as settled
  • history: optional rationale or decision context, only when it still helps future readers

Do not assume every project needs a separate home for every bucket. Some projects need explicit directories. Others only need better section boundaries.

Workflow

  1. Inspect the repository shape.
  • Start at the repo root.
  • Read AGENTS.md, ARCHITECTURE.md, README.md, and the docs index if they exist.
  • Load only the documentation files needed to understand the current documentation contract.
  • List the files that appear to define architecture, workflows, design, or planning.
  • Note whether the tree gives readers a legible path from overview docs into narrower topic or state-specific docs.
  1. Reconstruct the actual documentation contract.
  • Determine what the docs are trying to do today, even if the structure does not say it clearly.
  • Identify whether the repository is pre-bootstrap, partially implemented, or mature.
  • Note whether readers are likely to confuse target-state docs for implemented reality.
  1. Audit for confusion signals.
  • Look for present tense describing unbuilt systems.
  • Look for future-state architecture mixed into current-state onboarding docs.
  • Look for open questions embedded inside target docs.
  • Look for directory names or file names that hide the distinction between current and planned.
  • Look for trees where siblings mix overview docs, deep implementation notes, and planning artifacts without a visible organizing principle.
  • Look for directories that do not tell a reader what gets more specific as they descend.
  • Look for repeated rationale that does not help readers act.
  1. Classify each important doc.
  • For each important file, decide whether it is mainly current, target, question, or mixed.
  • If mixed, identify whether the mix is acceptable or whether it needs to be split.
  1. Recommend the lightest effective structure.
  • Choose the simplest methodology that makes the distinctions clear.
  • Possible outcomes include:

- improved section conventions only - time-horizon directories such as current/, target/, and questions/ - topic directories with explicit state sections inside each file - a hybrid, if the repository is large enough to justify it

  • Recommend where overview or index documents should live so each branch of the tree makes sense before readers open leaf files.
  • Explain why the chosen shape fits this specific repository.
  1. Propose writing rules.
  • Recommend tense and wording rules that keep the distinction visible in the body of the docs.
  • Recommend where unresolved material should live.
  • Recommend how much rationale should be retained.
  1. End with concrete next actions.
  • Name the smallest set of doc moves, renames, rewrites, or index changes that would materially reduce confusion.

Evaluation Heuristics

Use these heuristics when judging documentation quality:

Strong signals

  • A new reader can answer "what exists now?" in one or two clicks.
  • A new reader can answer "what are we trying to build?" separately.
  • A new reader can understand the documentation structure by scanning the tree and the first doc at each branch.
  • Parent docs give orientation, while child docs add detail rather than repeating or surprising.
  • Unresolved items are clearly marked and isolated.
  • Architecture docs do not quietly mix implemented behavior and aspirations.
  • The top-level docs tell readers where to start.

Risk signals

  • The same file alternates between current implementation facts and future design without headings.
  • The repository says it is early-stage, but architecture docs read as if the system already exists.
  • "Open questions" docs contain settled decisions.
  • The folder layout implies stable categories, but the writing inside ignores them.
  • The tree forces readers to open several files before they can tell how docs are organized.
  • Directory depth increases, but scope does not narrow predictably.
  • Metadata or frontmatter carries the truth distinction, but the visible prose does not.

Recommended Methodologies

Do not prescribe these mechanically. Pick one based on the repository.

Section-first

Use when:

  • the repository has few docs
  • splitting files would add more friction than clarity
  • a topic naturally contains both current and target information

Approach:

  • keep topic docs
  • add explicit sections such as Current State, Target State, and Open Questions
  • ensure the file opens with enough orientation that a reader can decide whether to keep reading or move deeper elsewhere

Trade-offs:

  • low directory churn
  • easy to adopt incrementally
  • depends on disciplined writing inside each file

Time-horizon-first

Use when:

  • readers are repeatedly confusing present state with planned state
  • the repo is pre-bootstrap or early-stage
  • there are already enough docs to justify structural separation

Approach:

  • use directories such as docs/current/, docs/target/, and docs/questions/
  • add short index or overview docs where readers need help choosing the right branch

Trade-offs:

  • very clear for new readers
  • can split one topic across several files
  • usually the best default for planning-heavy repositories

Topic-first with explicit status inside docs

Use when:

  • the project is mature
  • readers mostly navigate by subsystem rather than by timeline
  • implementation and design need to live close together by topic

Approach:

  • organize by topic such as architecture/, database/, and workflows/
  • require visible section boundaries in mixed docs
  • keep parent topic docs lightweight and directional, with deeper files carrying the narrower detail

Trade-offs:

  • scales well with system complexity
  • weaker by default at separating current and planned states
  • needs stronger writing discipline

Writing Rules

Recommend rules like these when they fit:

  • Use present tense only for implemented reality.
  • Use will, should, target, or equivalent language for intended state.
  • Keep unresolved items in a clearly marked place rather than buried in target docs.
  • At each directory level, make the scope narrower and more specific than the parent.
  • Put orientation near branch points and detail in leaf docs.
  • Keep rationale short and local unless the project truly benefits from decision-history documents.
  • Prefer visible headings over hidden metadata for state distinctions.

Output Shape

Use this shape unless the user asks for something else:

Overall Assessment

  • State whether the repository has low, medium, or high confusion risk.

What Readers Will Misread

  • Name the specific files or patterns likely to confuse new readers.

Best-Fit Methodology

  • Recommend the documentation methodology that best fits this repository.
  • Explain what it gains and what it gives up.

Reader Path

  • Explain whether the current tree supports progressive disclosure.
  • Recommend the smallest changes that make the navigation path legible.

Concrete Changes

  • List the smallest edits that would materially improve clarity.

Writing Rules

  • Give 3 to 6 practical rules the team can keep following.

Communication Rules

  • Prioritize clarity over doctrine.
  • Avoid acting like one documentation pattern is universally correct.
  • Tie every recommendation to the repository's current size, maturity, and risk of confusion.
  • Treat progressive disclosure as a reader-comprehension tool, not a mandate for more folders.
  • If the current structure is already good enough, say so directly and suggest only minimal refinements.
  • If a proposed structure adds maintenance cost, call that out explicitly.

Non-Goals

Do not:

  • turn the review into a generic style-guide lecture
  • insist on ADRs or frontmatter status fields by default
  • recommend heavyweight process without evidence the project needs it
  • confuse "more folders" with "better boundaries"
  • assume progressive disclosure always means deeper nesting

Example Triggers

  • "Check whether these docs confuse current state and future state."
  • "Audit this repository's docs organization."
  • "Do these docs make it obvious what exists now?"
  • "Should this project use current/ and target/ directories?"
  • "Find places where design docs read like implemented reality."
  • "Does this documentation tree support progressive disclosure?"
  • "Can a reader understand this docs structure just by looking at the tree?"

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

32.36%
按下载量换算39

Claude

30.87%
按下载量换算37

Cursor

19.84%
按下载量换算24

Gemini CLI

8.96%
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