Token导航 LogoToken导航TokenDH.com
研究检索只读clawhub未标认证来源可访问clear审计通过

draftdraft 搜索

Agent Skill

draft 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

10,783

周安装

432

GitHub Stars

公开资料未说明

下载量

3,491
OpenClaw

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

MIT-0

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:draft(draft 搜索)
来源仓库:https://github.com/ethagent/draft
安装命令:
openclaw skills install draft
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 OpenClaw 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。

ClawHubOpenClaw
openclaw skills install draft

简介

Draft 搜索用于查找、检索和筛选相关信息。

  • 适合根据关键词或任务场景快速定位候选结果。适用宿主包括 OpenClaw,接入前应确认版本、权限和运行环境要求。
  • 可作为专业初稿生成器,支持电子邮件、报告等多种文体。
  • 安装前建议确认是否会触发联网或文件读写操作。
  • draft 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

name
draft
description
>

Draft — First Draft Generator

What This Skill Does

Eliminates the blank page. Takes any input — a topic, bullet points, a messy brain dump, a one-line description — and produces a complete, polished first draft ready for immediate use or light editing.

Core Principle

A good first draft is not a perfect document. It is a complete document. It gives the user something real to react to, edit, and send — instead of a cursor blinking on an empty page.

Workflow

Step 1: Classify the Request

DOCUMENT_TYPES = {
  "email":        { structure: ["subject","greeting","context","ask","next_step","sign_off"], length: "50-300 words" },
  "report":       { structure: ["exec_summary","background","findings","analysis","recommendations"], length: "500-2000 words" },
  "article":      { structure: ["hook","context","body","conclusion","cta"], length: "600-1500 words" },
  "proposal":     { structure: ["problem","solution","methodology","timeline","investment","next_steps"], length: "400-1200 words" },
  "memo":         { structure: ["to/from/date/re","purpose","background","action"], length: "150-400 words" },
  "cover_letter": { structure: ["hook","why_them","why_me","evidence","ask"], length: "250-400 words" },
  "readme":       { structure: ["title","description","install","usage","examples","contributing"], length: "300-800 words" },
  "social_post":  { structure: ["hook","value","cta"], length: "50-500 words depending on platform" }
}

If document type is ambiguous, infer from context. Only ask if it materially changes the output.

Step 2: Extract the Brief

brief = {
  document_type:  classify(user_input),
  purpose:        what_should_this_accomplish(),
  audience:       who_will_read_this(),
  key_points:     what_must_be_included(),
  tone:           formal | professional | conversational | urgent | warm | bold,
  length:         short | medium | long,
  constraints:    deadlines, word_limits, sensitive_topics
}

Inference rules:

  • Casual writing → conversational tone
  • Mentions company or client → professional tone
  • Says "quick" or "short" → brevity is priority
  • Pastes bullet points → those ARE the key points, preserve all of them
  • No audience specified → infer from document type

Step 3: Write the Draft

Universal writing rules:

  • First sentence must earn attention or clearly state purpose
  • Every paragraph has one job — cut sentences that do not serve it
  • Active voice. Specific nouns. Strong verbs.
  • End with clarity: what happens next, what is being asked, what the reader should feel

Anti-patterns to eliminate:

  • "I hope this email finds you well" → delete
  • "In today's fast-paced world" → delete
  • "As per my previous email" → rewrite as "Following up on [specific thing]"
  • Paragraphs longer than 5 lines in emails
  • Burying the ask in paragraph 3

Step 4: Deliver and Offer Adjustments

Present the complete draft. Then offer exactly three targeted options:

Draft complete. Options if you want to refine:
1. [Specific tone adjustment relevant to this draft]
2. [Specific length adjustment relevant to this draft]
3. [Specific content addition relevant to this draft]

Or tell me what to change and I will revise immediately.

Format Rules by Document Type

  • Emails: plain text with Subject line labeled. No markdown inside the body.
  • Reports/Articles: markdown headers. Output ready to paste into a doc.
  • READMEs: full markdown with code blocks.
  • Social posts: output in a code block for easy copying. Include character count for Twitter/X.
  • Proposals: structured sections with headers. Numbers for timelines and pricing wherever possible.

Tone Calibration

TONE_GUIDE = {
  "formal":            "No contractions. Full sentences. Titles used.",
  "professional":      "Contractions OK. Direct. Clear. Respectful but not stiff.",
  "professional-warm": "Professional baseline plus genuine warmth. Human, not corporate.",
  "conversational":    "Write like a smart person talking. Short sentences. Real words.",
  "urgent":            "Lead with the urgency. Short paragraphs. Clear ask. No preamble.",
  "bold":              "Strong opinions. Active voice. No hedging. Confident assertions."
}

Quality Check Before Delivering

  • [ ] First sentence is strong — not a throat-clear
  • [ ] Every required piece of information from the brief is included
  • [ ] No placeholder text remaining
  • [ ] Tone matches the request and audience
  • [ ] Length is appropriate for document type
  • [ ] The ask or purpose is unmistakably clear
  • [ ] Draft is complete — not an outline, not bullets, a real document

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

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

补充不同宿主或平台的使用分布数据

能力 5

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

OpenClaw

72.78%
按下载量换算2,541

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

未展示

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