convez.ai
SkillAutomaçãoIntermediário
Fallme

Claude RPA Images

Fallme/claude-rpa-images

Download em lote inteligente de imagens com IA e automação de navegador

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill de automação que identifica e baixa em lote imagens de conteúdo de páginas web, distinguindo automaticamente entre imagens principais (wallpapers, produtos, arte) e elementos secundários (anúncios, ícones, UI). Implementa workflow de três passos com retry inteligente para garantir alta taxa de conclusão.

Instalação

$npx convez-ai Fallme/claude-rpa-images

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Extrair wallpapers de galerias web mantendo ordem
  • Baixar catálogos de imagens de produtos em massa
  • Fazer scraping de mangá/quadrinhos preservando sequência de páginas

Veredito

Ferramenta especializada em RPA para Claude Code com algoritmo de classificação robusto (5 camadas) e workflow resiliente de retry. Útil para casos específicos de extração de imagens, mas exige configuração manual e conhecimento técnico de URLs/DOM.

README

A tradução para o português deste README ainda está na fila.

Claude RPA Images

智能批量图片下载 — Claude Code 的 AI 浏览器自动化 Skill

Smart batch image download for Claude Code — AI-powered browser automation skill


简介

/rpa-images 是一个 Claude Code Skill,可以从任意网页智能识别并批量下载主图内容。自动区分内容图片(壁纸、产品图、漫画)与广告、图标、UI 元素,支持三轮下载保证高完成率。

/rpa-images is a Claude Code Skill that intelligently identifies and batch-downloads main content images from any webpage. It automatically distinguishes content images (wallpapers, product photos, artwork) from thumbnails, icons, ads, and UI elements. A three-pass download workflow ensures 95-100% completion.

核心功能 / Features

  • 智能图片分类 — 5层筛选算法:硬性尺寸过滤 → DOM结构分析 → 尺寸分布检测(广告标准尺寸728x90/300x250等)→ 链接上下文 → AI兜底判断

  • Smart Image Classification — 5-layer filtering: hard size filter → DOM structure → size distribution (ad standard sizes 728x90/300x250) → link context → AI fallback

  • 三轮下载工作流 — Pass 1 按序下载(2-3s间隔) → Pass 2 缺口补填(5s间隔) → Pass 3 最终复查(8s间隔,最多3轮重试)

  • Three-Pass Download Workflow — Pass 1 ordered download (2-3s delay) → Pass 2 gap-fill (5s delay) → Pass 3 final verify (8s delay, up to 3 retry rounds)

  • 顺序保持 — 漫画/漫画等连续内容保持DOM原始顺序,文件按页码命名(01.jpeg, 02.jpeg...)

  • Sequential Ordering — Comics/manga maintain original DOM order, files named by page number (01.jpeg, 02.jpeg...)

  • CDN容错 — 自动处理301重定向、SSL错误、速率限制,curl + Python urllib双引擎

  • CDN Resilience — Handles 301 redirects, SSL errors, rate limiting automatically with dual curl + Python urllib engines

  • 自动文件夹 — 使用网页 <title> 自动创建目标文件夹

  • Auto Folder — Creates target folder using the page <title> automatically

安装 / Installation

将 rpa-images 文件夹复制到 Claude Code 的 skills 目录:

Copy the rpa-images folder to your Claude Code skills directory:

# 从本仓库克隆 / Clone this repo
git clone https://github.com/Fallme/claude-rpa-images.git

# 复制到 skills 目录 / Copy to skills directory
cp -r claude-rpa-images/rpa-images ~/.claude/skills/

使用方法 / Usage

在 Claude Code 中直接调用:

Invoke directly in Claude Code:

/rpa-images https://example.com/wallpaper-page

支持可选参数 / Optional parameters:

/rpa-images <url> [图片类型: wallpaper/product/photo/art] [最小尺寸过滤]

示例 / Examples

# 下载壁纸页面所有主图 / Download all main images from wallpaper page
/rpa-images https://example.com/wallpapers/12345

# 下载产品图(仅大于300px的)/ Download product images (only >300px)
/rpa-images https://example.com/product/67890 product 300

# 下载漫画(保持页码顺序)/ Download manga (preserve page order)
/rpa-images https://example.com/manga/chapter-1

下载工作流详解 / Download Workflow Details

Pass 1: 按序下载 / Ordered Download

  • 提取页面中所有内容图片URL,保持DOM原始顺序
  • Extract all content image URLs preserving DOM order
  • 每张图片间隔2-3秒下载
  • 2-3 second delay between downloads
  • 文件命名:01.jpeg, 02.jpeg, ...
  • 失败的URL记录到列表
  • Failed URLs logged to list

Pass 2: 缺口补填 / Gap Fill

  • 使用Python urllib重试Pass 1失败的URL
  • Retry failed URLs from Pass 1 with Python urllib
  • 间隔延长到5秒,添加完整浏览器请求头
  • 5 second delay, full browser headers
  • 自动处理301重定向(curl失败时)
  • Auto-follow 301 redirects (when curl fails)

Pass 3: 最终复查 / Final Verify

  • 对比已下载文件数与预期数量
  • Compare downloaded count vs expected
  • 仍然缺失的文件最多重试3轮,间隔8秒
  • Still-missing files retried up to 3 rounds, 8s delay
  • 生成完整报告,列出缺失页码
  • Generate completion report with missing page numbers

错误恢复 / Error Recovery

错误 / Error策略 / Strategy
HTTP 301 重定向 / redirect提取Location头,Python urllib自动跟随
HTTP 403添加Referer头,完整浏览器UA
SSL错误 / errorPython urllib + ssl._create_unverified_context()
速率限制 / rate limit渐进延迟:3s → 5s → 8s
懒加载 / lazy loading检查 data-* 属性,滚动触发加载
反爬虫 / anti-bot检测验证页面,暂停等待用户处理

技术架构 / Architecture

rpa-images/
└── SKILL.md          # Skill定义文件 / Skill definition file

Skill 采用 Markdown 格式定义,Claude Code 在执行时自动解析 pipeline 阶段和决策逻辑。

The skill is defined in Markdown format. Claude Code automatically parses pipeline stages and decision logic at runtime.

依赖 / Dependencies

  • Claude Code with Claude in Chrome 扩展 / extension (browser automation)
  • Python 3 (urllib, ssl — 标准库,无需额外安装 / standard library, no extra install)
  • curl (可选,Pass 1使用 / optional, used in Pass 1)

相关项目 / Related Projects

License

MIT

Ver o original no GitHub

Bom para

Relacionados