Token导航 LogoToken导航

Agent Skill · 研究检索

kimi-pdf(Kimi PDF 搜索)

kimi-pdf 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 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。

kimi-pdf 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,权限为操作浏览器;许可证信息还不明确,第三方扫描有未通过项,按安装量在研究检索类的 73,675 个 Skill 里排第 24,782。

总安装 3,927周安装 162GitHub Stars 146下载 1,283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:kimi-pdf(Kimi PDF 搜索)
来源仓库:https://github.com/thvroyal/kimi-skills
仓库路径:skills/kimi-pdf
安装命令:
npx skills add https://github.com/thvroyal/kimi-skills --skill kimi-pdf
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/thvroyal/kimi-skills --skill kimi-pdf

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。

kimi-pdf 用于查找、检索和筛选相关信息,适合在多种宿主环境中快速定位内容。适用于根据关键词或任务场景从来源线索中筛选信息的需求。通过 npx skills add 命令安装,需结合原始 README 核验具体用法。安装前建议确认权限范围、维护状态及是否触发联网或文件操作。

kimi-pdf 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/kimi-pdf

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/thvroyal/kimi-skills --skill kimi-pdf;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

    它需要操作浏览器,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

kimi-pdf 怎么安装?

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

kimi-pdf 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。

kimi-pdf 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

kimi-pdf 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。

kimi-pdf 用的人多吗?

公开数据是累计安装 3,927 次,近一周安装 162 次,下载 1,283 次,GitHub 146 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。

kimi-pdf 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

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。

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。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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
kimi-pdf

Kimi PDF 搜索 · 本页

3,927146操作浏览器未明确未通过

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

Route Selection

RouteTriggerRoute File
HTML (default)All PDF creation requestsroutes/html.md
LaTeXUser explicitly requests LaTeX,.tex, or Tectonicroutes/latex.md
ProcessWork with existing PDFs (extract, merge, fill forms, etc.)routes/process.md

Default to HTML. Only use LaTeX route when user explicitly requests it.

MANDATORY: Read Route File Before Implementation

Before implementation, you MUST:

  1. Determine the route (HTML / LaTeX / Process)
  2. Read the route file (routes/html.md, routes/latex.md, or routes/process.md)
  3. Only then proceed with implementation

This file (SKILL.md) contains constraints and principles. Route files contain how-to details.

Decision Rules

Route Selection
User SaysRoute
"Create a PDF", "Make a report", "Write a paper"HTML
"Use LaTeX", "Compile.tex", "Use Tectonic"LaTeX
"Extract text from PDF", "Merge these PDFs", "Fill this form"Process
Cover Style Selection (HTML Route)
ContextStyle
Academic paper, thesis, formal courseworkMinimal (white, centered, no decoration)
Reports, proposals, professional documentsDesigned (choose from style reference in html.md)
UncertainDefault to Designed — plain text cover = mediocre

Key principle: Cover background separates "acceptable" from "impressive". See html.md for 11 style options.

Citation Format Selection
Document LanguageFormat
ChineseGB/T 7714 (use [J][M][D] identifiers)
EnglishAPA
MixedChinese refs → GB/T 7714, English refs → APA

Quick Start

Use the unified CLI for all operations:

# Check environment (JSON output, exit code 0=ok, 2=missing deps)
/app/.kimi/skills/kimi-pdf/scripts/pdf.sh check

# Auto-fix missing dependencies (idempotent, safe to run multiple times)
/app/.kimi/skills/kimi-pdf/scripts/pdf.sh fix

# Convert HTML to PDF
/app/.kimi/skills/kimi-pdf/scripts/pdf.sh html input.html

# Compile LaTeX to PDF
/app/.kimi/skills/kimi-pdf/scripts/pdf.sh latex input.tex

Exit codes:

  • 0 = success
  • 1 = usage error
  • 2 = dependency missing (run pdf.sh fix)
  • 3 = runtime error

Dependencies by route:

  • HTML route: Node.js, Playwright, Chromium
  • Process route: Python 3, pikepdf, pdfplumber
  • LaTeX route: Tectonic

Core Constraints (Must Follow)

1. Output Language

Output language must match user's query language.

  • User writes in Chinese → PDF content in Chinese
  • User writes in English → PDF content in English
  • User explicitly specifies language → Follow user's specification

2. Word Count and Page Constraints

  • Strictly follow user-specified word/page count requirements
  • Do not arbitrarily inflate content length

3. Citation and Search Standards

CRITICAL: Search Before Writing

DO NOT fabricate information. When in doubt, SEARCH.

If content involves ANY of these, you MUST search FIRST before writing:

  • Statistics, numbers, percentages, rankings
  • Policies, regulations, laws, standards
  • Academic research, theories, methodologies
  • Current events, recent developments
  • Anything you're not 100% certain about
When Search is Required
ScenarioSearch?Notes
Statistics, dataRequirede.g., "2024 employment rate"
Policies, regulationsRequirede.g., "startup subsidies"
Research, papersRequirede.g., "effectiveness of method X"
Time-sensitive contentRequiredInformation after knowledge cutoff
Uncertain factsRequiredIf unsure, always search
Common knowledgeNot needede.g., "water boils at 100°C"

Search workflow:

  1. Identify facts/data requiring verification
  2. Search for authentic sources
  3. If results insufficient, iterate search until reliable info obtained
  4. Include real sources in references
  5. If search fails repeatedly, tell the user instead of making up data
Citations Must Be Real

Fabricating references is prohibited. All citations must have:

  • Correct author/institution names
  • Accurate titles
  • Verifiable year, journal/source
Cross-references (Must Be Clickable)
As shown in <a href="#fig-1-1">Figure 1-1</a>...
From <a href="#eq-2-1">Equation (2-1)</a>...
See <a href="#sec3">Section 3</a>...

Note: id must be placed at container top (see CSS Counters section in html.md).


Content Quality Constraints

1. Word/Page Count Constraints

Must strictly follow user-specified word or page count requirements:

User RequestExecution Standard
Specific word count (e.g., "3000 words")Within ±20%, i.e., 2400-3600 words
Specific page count (e.g., "5 pages")Exactly equal, last page may be partial
Word count range (e.g., "2000-3000 words")Must fall within range
No explicit requirementInfer reasonably by document type; prefer thorough over superficial
Minimum specified (e.g., "more than 5000 words")No more than 2x, i.e., 5000-10000 words

Prohibited behaviors:

  • Arbitrarily shortening content ("concise" is not an excuse)
  • Padding pages with excessive bullet lists (maintain high information density)
  • Exceeding twice the user's requested length

Special case - Resume/CV:

  • Resume should be 1 page unless user specifies otherwise
  • Use compact margins: margin: 1.5cm

2. Outline Adherence (Mandatory)

When user provides outline:

  • Strictly follow the user-provided outline structure
  • Section titles must match outline (minor wording adjustments OK, no level/order changes)
  • Do not add or remove sections arbitrarily
  • If outline seems problematic, ask user first before modifying

When no user outline:

  • Use standard structures based on document type:

- Academic papers: IMRaD (Introduction-Methods-Results-Discussion) or Introduction-Literature Review-Methods-Results-Discussion-Conclusion - Business reports: Conclusion-first (Executive Summary → Detailed Analysis → Recommendations) - Technical docs: Overview → Principles → Usage → Examples → FAQ - Course assignments: Follow assignment structure requirements

  • Sections must have logical progression, no disconnects

Tech Stack Overview

RouteToolsPurpose
HTMLPlaywright + Paged.jsHTML → PDF conversion
HTMLKaTeX, MermaidMath formulas, diagrams
ProcesspikepdfForm filling, page operations, metadata
ProcesspdfplumberText and table extraction
ProcessLibreOfficeOffice → PDF conversion
LaTeXTectonicLaTeX → PDF compilation

平台分布

Codex

35.85%
按下载量换算460

Claude

32.32%
按下载量换算415

Cursor

19.82%
按下载量换算254

Gemini CLI

9.48%
按下载量换算122

安全审计

Gen Agent Trust Hub

未通过

Socket

可疑

Snyk

可疑

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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