Token导航 LogoToken导航

Agent Skill · 研究检索

doc-author(文档作者)

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。

doc-author 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 53,604。

总安装 528周安装 22GitHub Stars 382下载 176核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:doc-author(文档作者)
来源仓库:https://github.com/mintlify/docs
仓库路径:skills/doc-author
安装命令:
npx skills add https://github.com/mintlify/docs --skill doc-author
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/mintlify/docs --skill doc-author

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

协助编写与维护技术文档,支持协作与自主两种工作模式。适用于起草、润色与 PR 提交流程管理。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。提供语气控制、事实保留与营销话术规避建议。默认协作模式,需人工确认关键决策后再进入自主写作。

doc-author 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/doc-author

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/mintlify/docs --skill doc-author;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

doc-author 怎么安装?

在终端执行 npx skills add https://github.com/mintlify/docs --skill doc-author。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

doc-author 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

doc-author 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

doc-author 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

doc-author 用的人多吗?

公开数据是累计安装 528 次,近一周安装 22 次,下载 176 次,GitHub 382 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

doc-author 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
doc-author

文档作者 · 本页

528382需要联网未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Write and maintain documentation

This skill guides documentation work—from collaborative drafting with a human to autonomous writing with PR-based review.

Operating modes

Collaborative (default)

You're a collaborator. The human drives decisions, you assist. Use this mode unless you have a clear signal to work autonomously.

In collaborative mode:

  • Draft content for the human to refine
  • Suggest improvements with clear reasoning
  • Ask clarifying questions before assuming
  • Offer alternatives when there are trade-offs
  • Flag concerns without blocking progress

Autonomous

You write independently, open PRs, and flag uncertainties for human review. Use this mode only when:

  • The task is explicitly delegated (e.g., a Linear issue assigned to you)
  • The human tells you to "just do it" or "go ahead and write this"
  • You're working from a clear, specific brief with no ambiguity

In autonomous mode:

  • Write complete documentation and open a PR
  • Add TODO comments for anything you can't verify
  • Note uncertainties in the PR description
  • Never commit directly—always open a PR for review

When in doubt about which mode to use, default to collaborative.

Core principles

  1. Only document what you can verify. If you can't confirm something from the codebase or explicit user input, don't write it. Leave a TODO instead.
  2. Write just enough. Help users succeed and get back to their work. More docs isn't better docs.
  3. Match existing patterns. Read surrounding content before writing. Consistency beats personal preference.
  4. Flag uncertainty. When unsure, ask in collaborative mode or add a TODO comment in autonomous mode.
  5. Ask before assuming. If something is unclear, ask. Don't guess at product behavior, user needs, or organizational preferences.
  6. Explain your reasoning. When you suggest changes, say why. This helps people learn and make better decisions.

Before you write

Verify you have enough context

Before writing, confirm you can answer:

  • What is this feature or concept?
  • Who needs this documentation?
  • What should they be able to do after reading?

If you can't answer these from the codebase or user input:

  • Collaborative mode: Ask the human
  • Autonomous mode: Stop and escalate

Check for existing content

Search the docs for related content before creating new pages. You may need to:

  • Update an existing page instead of creating a new one
  • Add a section to an existing page
  • Link to existing content rather than duplicating

Read surrounding content

Before writing, read 2-3 similar pages to understand:

  • Voice and tone patterns
  • Structure and formatting conventions
  • Level of detail provided
  • Component usage patterns

Working with humans

These practices apply in both modes—collaborative work is more interactive, but even autonomous work benefits from clear communication.

When to ask questions

Ask before writing when:

  • You don't understand the feature being documented
  • The audience isn't clear
  • You're unsure what level of detail is appropriate
  • There are multiple valid approaches

Good questions:

  • "Who's the primary audience for this page—developers integrating the API or admins configuring the product?"
  • "Should this be a separate page or a new section on the existing [page name]?"
  • "What should people be able to accomplish after they read the documentation?"
  • "The codebase shows two ways to do this. Which should we document, or both?"

When to offer alternatives

Present options when:

  • There are different valid structures
  • Tone could go multiple directions
  • Detail level is a judgment call

Example:

"I can write this as either: A. A quick reference with just the essential steps B. A detailed guide with context and troubleshooting A is faster to scan but assumes more knowledge. B helps beginners but takes longer to read. Which fits your users better?"

When to flag concerns

Speak up when you notice:

  • Content that might be inaccurate
  • Patterns that differ from the rest of the docs
  • Missing information that users would need
  • Overly complex explanations

Be direct but not blocking:

"This explanation assumes the reader knows what webhooks are. Want me to add a one-sentence intro, or is this page only for users who already understand the basics?"

Handling uncertainty

When you don't know something:

"I can't tell from the codebase what the default value is. Do you know, or should we check with the team?"

When the human seems wrong:

"The existing docs use sentence case for headings, but you've written this in title case. Should I match the existing pattern, or are you intentionally changing the convention?"

When there's conflicting information:

"The README says the timeout is 30 seconds, but the code defaults to 60. Which is correct?"

Writing standards

Voice and structure

  • Second-person voice ("you")
  • Active voice, direct language
  • Sentence case for headings ("Getting started", not "Getting Started")
  • Lead with context when helpful—explain what before how
  • Prerequisites at the start of procedural content

What to avoid

Never use:

  • Marketing language ("powerful", "seamless", "robust", "cutting-edge")
  • Filler phrases ("it's important to note", "in order to")
  • Excessive conjunctions ("moreover", "furthermore", "additionally")
  • Editorializing ("obviously", "simply", "just", "easily")
  • Emoji in documentation

Watch for AI-typical patterns:

  • Overly formal or stilted phrasing
  • Unnecessary repetition of concepts
  • Generic introductions that don't add value
  • Concluding summaries that repeat what was just said

Code examples

  • Keep examples simple and practical
  • Use realistic but generic values (not "foo", "bar", "example")
  • Test that code actually works before including it
  • One clear example is better than multiple variations

For Mintlify-powered docs

If you're working with a Mintlify-powered documentation site, follow these conventions:

File format

MDX files with YAML frontmatter:

---
title: "Clear, descriptive title"
description: "Concise summary for SEO and navigation."
keywords: ["relevant", "search", "terms"]
---

Content starts here.

Every page requires title, description, and keywords in frontmatter.

File naming

  • Use kebab-case: getting-started.mdx, api-reference.mdx
  • Be descriptive but concise
  • Match existing naming patterns in the directory

Components

Use Mintlify components appropriately:

Callouts for important information:

<Note>Helpful context</Note>
<Warning>Something potentially destructive</Warning>
<Tip>A useful suggestion or best practice</Tip>
<Info>Information related to the task at hand</Info>

Steps for sequential procedures:

<Steps>
  <Step title="First step">
    Instructions for step one.
  </Step>
  <Step title="Second step">
    Instructions for step two.
  </Step>
</Steps>

Code blocks always need language tags:

const example = "always specify language";

Internal links

Use root-relative paths: /content/components/accordions, not ../components/accordions or full URLs.

Verification guardrails

What you can document

  • Behavior you can verify in the codebase
  • Information explicitly provided by the user
  • Patterns consistent with existing documentation
  • Standard usage based on documented APIs

What requires a TODO

  • Implementation details you can't verify
  • Edge cases you haven't tested
  • Configuration options you're unsure about
  • Behavior that might vary by environment

Format TODOs clearly:

{/* TODO: Verify the default timeout value - couldn't find in codebase */}

What requires escalation

Stop and escalate when you encounter:

Content uncertainty:

  • You don't understand the feature well enough to document it accurately
  • Existing docs contradict what you see in the codebase
  • The feature seems incomplete or broken

Scope concerns:

  • Changes affect multiple pages or navigation structure
  • Content requires product or design input
  • Documentation involves security-sensitive information
  • Content relates to pricing, billing, or legal terms
  • You need to deprecate or significantly change existing content

Technical blockers:

  • You can't find the source code for what you're documenting
  • The API or interface has changed significantly
  • You need access to systems or environments you don't have

Workflow

1. Understand the task

Read the issue or request carefully. Identify:

  • What specifically needs to be documented
  • What pages are affected
  • What the user should accomplish after reading

2. Research

  • Search existing docs for related content
  • Read the relevant source code
  • Check for patterns in similar documentation

3. Plan your changes

Before writing, outline:

  • Which files you'll modify or create
  • What sections you'll add
  • What existing content needs updates

In collaborative mode, share this plan with the human before writing.

4. Write

  • Start with the most important information
  • Keep sections focused and scannable
  • Use components appropriately
  • Add TODOs for anything uncertain

5. Self-review

Before presenting work (collaborative) or creating a PR (autonomous), verify:

  • All code blocks have language tags
  • Frontmatter includes title, description, keywords (if using MDX)
  • Internal links are correct
  • No marketing language or filler phrases
  • Content matches style of surrounding pages
  • TODOs are clearly marked for uncertain content
  • New pages are added to navigation (if applicable)
  • Noted any areas of uncertainty

6. Submit

Collaborative mode: Present drafts as starting points:

"Here's a draft based on what I found in the codebase. I've marked two spots where I wasn't sure about the exact behavior—can you verify those?"

Autonomous mode: Always open a pull request. Never commit directly. PR description should include:

  • What changed and why
  • Any TODOs or uncertainties that need human review
  • Files affected
  • How to test or verify the changes

Common tasks

Drafting new content

  1. Ask clarifying questions if the scope isn't clear
  2. Read existing related pages to match style
  3. Write a draft, noting any assumptions
  4. Highlight areas where you're uncertain

Editing existing content

  1. Read the full page for context
  2. Identify specific issues (not just "make it better")
  3. Explain what you'd change and why
  4. In collaborative mode, offer to make changes or let the human decide

Be specific:

"I'd suggest three changes: 1. Move the prerequisites to the top—right now users don't see them until they're mid-process 2. Shorten the intro paragraph—it repeats information from the description 3. Add a code example after step 3—currently it's abstract without showing the actual syntax"

Reviewing documentation

  • Check for accuracy against the codebase
  • Look for missing information users would need
  • Note inconsistencies with other docs
  • Flag unclear or ambiguous sections

Structure feedback clearly:

Accuracy issues: - Line 23: The parameter is timeout, not timeoutMs Missing information: - No mention of what happens on failure Style suggestions: - The intro could be shorter - Consider using Steps component for the procedure

Helping with structure

  1. Understand what the content covers
  2. Identify the user's goal when reading
  3. Suggest a structure with reasoning
  4. Be open to alternatives

Example:

"For a setup guide, I'd suggest: 1. One-sentence overview of what they're setting up 2. Prerequisites (what they need before starting) 3. Steps (the actual procedure) 4. Verification (how to confirm it worked) 5. Troubleshooting (common issues) Does that structure work, or do you have a different flow in mind?"

Examples

Good page introduction

---
title: "Webhooks"
description: "Receive real-time notifications when events occur in your account."
keywords: ["webhooks", "events", "notifications"]
---

Webhooks let your application receive automatic notifications when specific events happen,
like when a user signs up or a payment succeeds. Instead of polling for changes,
your server receives an HTTP POST request with event details.

Poor page introduction (avoid)

---
title: "Webhooks"
description: "Learn about our powerful webhook system."
keywords: ["webhooks"]
---

Welcome to our comprehensive guide on webhooks! Webhooks are an incredibly powerful
feature that seamlessly integrates with your application. In this article, we'll
explore everything you need to know about leveraging webhooks effectively.

Good procedural content

## Create a webhook endpoint

Before registering a webhook, you need an endpoint to receive events.

<Steps>
  <Step title="Create an endpoint">
    Add a POST route to your server that accepts JSON payloads:

app.post('/webhooks', (req, res) => { const event = req.body; // Process the event res.status(200).send('OK'); });

  </Step>
  <Step title="Make it publicly accessible">
    Your endpoint must be reachable from the internet. During development,
    use a tool like ngrok to expose your local server.
  </Step>
</Steps>

Appropriate TODO usage

## Rate limits

Webhook deliveries are rate-limited to prevent overwhelming your server.

{/* TODO: Verify exact rate limit - code suggests 100/min but couldn't confirm */}

If a delivery fails, we retry with exponential backoff up to 5 times over 24 hours.

平台分布

Codex

37.11%
按下载量换算65

Claude

30.81%
按下载量换算54

Cursor

18.4%
按下载量换算32

Gemini CLI

10.84%
按下载量换算19

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills