apify（apify 搜索）

Customization

Before executing, check for user customizations at: ~/.claude/PAI/USER/SKILLCUSTOMIZATIONS/Apify/

If this directory exists, load and apply any PREFERENCES.md, configurations, or resources found there. These override default behavior. If the directory does not exist, proceed with skill defaults.

🚨 MANDATORY: Voice Notification (REQUIRED BEFORE ANY ACTION)

You MUST send this notification BEFORE doing anything else when this skill is invoked.

1. Send voice notification: curl -s -X POST http://localhost:8888/notify \ -H "Content-Type: application/json" \ -d '{"message": "Running the WORKFLOWNAME workflow in the Apify skill to ACTION"}' \ > /dev/null 2>&1 &
2. Output text notification: Running the **WorkflowName** workflow in the **Apify** skill to ACTION...

This is not optional. Execute this curl command immediately upon skill invocation.

Apify - Social Media & Web Scraping

Direct TypeScript access to 9 popular Apify actors with 99% token savings.

🔌 File-Based MCP

This skill is a file-based MCP - a code-first API wrapper that replaces token-heavy MCP protocol calls.

Why file-based? Filter data in code BEFORE returning to model context = 97.5% token savings.

Architecture: See ~/.claude/PAI/DOCUMENTATION/FileBasedMCPs.md

🎯 Overview

Direct TypeScript access to the 9 most popular Apify actors without MCP overhead. Filter and transform data in code BEFORE it reaches the model context.

📊 Available Actors

Social Media (5 platforms)

• Instagram (145k users, 4.60★) - Profiles, posts, hashtags, comments
• LinkedIn (26k users, 4.10★) - Profiles, jobs, posts
• TikTok (90k users, 4.61★) - Profiles, videos, hashtags, comments
• YouTube (40k users, 4.40★) - Channels, videos, comments, search
• Facebook (35k users, 4.56★) - Posts, groups, comments

Business & Lead Generation

• Google Maps (198k users, 4.76★) - HIGHEST VALUE!

- Search businesses, extract contacts, reviews, images - Perfect for lead generation

E-commerce

• Amazon (8k users, 4.97★) - Products, reviews, pricing

Web Scraping

• Web Scraper (94k users, 4.39★) - General-purpose, works with ANY website

🚀 Quick Start

Basic Usage Pattern

import { scrapeInstagramProfile, searchGoogleMaps } from 'actors'

// 1. Call the actor wrapper
const profile = await scrapeInstagramProfile({
  username: 'target_username',
  maxPosts: 50
})

// 2. Filter in code - BEFORE data reaches model!
const viral = profile.latestPosts?.filter(p => p.likesCount > 10000)

// 3. Only filtered results reach model context
console.log(viral) // ~10 posts instead of 50

📚 Examples by Use Case

Social Media Monitoring

Instagram - Track engagement:

import { scrapeInstagramProfile, scrapeInstagramPosts } from 'actors'

// Get profile with recent posts
const profile = await scrapeInstagramProfile({
  username: 'competitor',
  maxPosts: 100
})

// Filter in code - only high-performing posts from last 30 days
const thirtyDaysAgo = Date.now() - (30 * 24 * 60 * 60 * 1000)
const topRecent = profile.latestPosts
  ?.filter(p =>
    new Date(p.timestamp).getTime() > thirtyDaysAgo &&
    p.likesCount > 5000
  )
  .sort((a, b) => b.likesCount - a.likesCount)
  .slice(0, 10)

// Only 10 posts reach model instead of 100!

LinkedIn - Job search:

import { searchLinkedInJobs } from 'actors'

const jobs = await searchLinkedInJobs({
  keywords: 'AI engineer',
  location: 'San Francisco',
  remote: true,
  maxResults: 200
})

// Filter in code - only senior roles at well-funded startups
const topJobs = jobs.filter(j =>
  j.seniority?.includes('Senior') &&
  parseInt(j.applicants || '0') > 50
)

TikTok - Trend analysis:

import { scrapeTikTokHashtag } from 'actors'

const videos = await scrapeTikTokHashtag({
  hashtag: 'ai',
  maxResults: 500
})

// Filter in code - only viral content
const viral = videos
  .filter(v => v.playCount > 1000000)
  .sort((a, b) => b.playCount - a.playCount)
  .slice(0, 20)

Lead Generation (Business Intelligence)

Google Maps - Local business leads:

import { searchGoogleMaps } from 'actors'

// Search with contact info extraction
const places = await searchGoogleMaps({
  query: 'restaurants in Austin',
  maxResults: 500,
  includeReviews: true,
  maxReviewsPerPlace: 20,
  scrapeContactInfo: true // Extracts emails from websites!
})

// Filter in code - only highly-rated with email/phone
const qualifiedLeads = places
  .filter(p =>
    p.rating >= 4.5 &&
    p.reviewsCount >= 100 &&
    (p.email || p.phone)
  )
  .map(p => ({
    name: p.name,
    rating: p.rating,
    reviews: p.reviewsCount,
    email: p.email,
    phone: p.phone,
    website: p.website,
    address: p.address
  }))

// Export leads - only qualified results!
console.log(`Found ${qualifiedLeads.length} qualified leads`)

Google Maps - Review sentiment analysis:

import { scrapeGoogleMapsReviews } from 'actors'

const reviews = await scrapeGoogleMapsReviews({
  placeUrl: 'https://maps.google.com/maps?cid=12345',
  maxResults: 1000
})

// Filter in code - analyze sentiment by rating
const recentNegative = reviews
  .filter(r => {
    const thirtyDaysAgo = Date.now() - (30 * 24 * 60 * 60 * 1000)
    return (
      r.rating <= 2 &&
      new Date(r.publishedAtDate).getTime() > thirtyDaysAgo &&
      r.text.length > 50
    )
  })

// Identify common complaints
const complaints = recentNegative.map(r => r.text)

E-commerce & Competitive Intelligence

Amazon - Price monitoring:

import { scrapeAmazonProduct } from 'actors'

const product = await scrapeAmazonProduct({
  productUrl: 'https://www.amazon.com/dp/B08L5VT894',
  includeReviews: true,
  maxReviews: 200
})

// Filter in code - only recent negative reviews
const recentNegative = product.reviews
  ?.filter(r => {
    const weekAgo = Date.now() - (7 * 24 * 60 * 60 * 1000)
    return (
      r.rating <= 2 &&
      new Date(r.date).getTime() > weekAgo
    )
  })

console.log(`Price: $${product.price}`)
console.log(`Rating: ${product.rating}/5`)
console.log(`Recent issues: ${recentNegative?.length} complaints`)

Custom Web Scraping

Any Website - Custom extraction:

import { scrapeWebsite } from 'actors'

const products = await scrapeWebsite({
  startUrls: ['https://example.com/products'],
  linkSelector: 'a.product-link',
  maxPagesPerCrawl: 100,
  pageFunction: `
    async function pageFunction(context) {
      const { request, $, log } = context

      return {
        url: request.url,
        title: $('h1.product-title').text(),
        price: $('span.price').text(),
        inStock: $('.in-stock').length > 0,
        description: $('.description').text()
      }
    }
  `
})

// Filter in code - only available products under $100
const affordable = products.filter(p =>
  p.inStock &&
  parseFloat(p.price.replace('$', '')) < 100
)

🎨 Advanced Patterns

Pattern 1: Multi-Platform Social Listening

import {
  scrapeInstagramHashtag,
  scrapeTikTokHashtag,
  searchYouTube
} from 'actors'

// Run all platforms in parallel
const [instagramPosts, tiktokVideos, youtubeVideos] = await Promise.all([
  scrapeInstagramHashtag({ hashtag: 'ai', maxResults: 100 }),
  scrapeTikTokHashtag({ hashtag: 'ai', maxResults: 100 }),
  searchYouTube({ query: '#ai', maxResults: 100 })
])

// Combine and filter - only viral content across all platforms
const allViral = [
  ...instagramPosts.filter(p => p.likesCount > 10000),
  ...tiktokVideos.filter(v => v.playCount > 100000),
  ...youtubeVideos.filter(v => v.viewsCount > 50000)
]

console.log(`Found ${allViral.length} viral posts across 3 platforms`)

Pattern 2: Lead Enrichment Pipeline

import { searchGoogleMaps, scrapeLinkedInProfile } from 'actors'

// 1. Find businesses on Google Maps
const restaurants = await searchGoogleMaps({
  query: 'restaurants in SF',
  maxResults: 100,
  scrapeContactInfo: true
})

// 2. Filter for qualified leads
const qualified = restaurants.filter(r =>
  r.rating >= 4.5 &&
  r.email &&
  r.reviewsCount >= 50
)

// 3. Enrich with LinkedIn data (if available)
const enriched = await Promise.all(
  qualified.map(async (restaurant) => {
    // Try to find LinkedIn company page
    // ... additional enrichment logic
    return restaurant
  })
)

Pattern 3: Competitive Analysis Dashboard

import {
  scrapeInstagramProfile,
  scrapeYouTubeChannel,
  scrapeTikTokProfile
} from 'actors'

async function analyzeCompetitor(username: string) {
  // Gather data from all platforms
  const [instagram, youtube, tiktok] = await Promise.all([
    scrapeInstagramProfile({ username, maxPosts: 30 }),
    scrapeYouTubeChannel({ channelUrl: `https://youtube.com/@${username}`, maxVideos: 30 }),
    scrapeTikTokProfile({ username, maxVideos: 30 })
  ])

  // Calculate engagement metrics in code
  return {
    username,
    instagram: {
      followers: instagram.followersCount,
      avgLikes: average(instagram.latestPosts?.map(p => p.likesCount) || []),
      engagementRate: calculateEngagement(instagram)
    },
    youtube: {
      subscribers: youtube.subscribersCount,
      avgViews: average(youtube.videos?.map(v => v.viewsCount) || [])
    },
    tiktok: {
      followers: tiktok.followersCount,
      avgPlays: average(tiktok.videos?.map(v => v.playCount) || [])
    }
  }
}

💰 Token Savings Calculator

Example: Instagram profile with 100 posts

MCP Approach:

1. search-actors → 1,000 tokens
2. call-actor → 1,000 tokens
3. get-actor-output → 50,000 tokens (100 unfiltered posts)
TOTAL: ~52,000 tokens

File-Based Approach:

const profile = await scrapeInstagramProfile({
  username: 'user',
  maxPosts: 100
})

// Filter in code - only top 10 posts
const top = profile.latestPosts
  ?.sort((a, b) => b.likesCount - a.likesCount)
  .slice(0, 10)

// TOTAL: ~500 tokens (only 10 filtered posts reach model)

Savings: 99% reduction (52,000 → 500 tokens)

🔧 Actor Reference

Social Media

Instagram

• scrapeInstagramProfile(input) - Profile + posts
• scrapeInstagramPosts(input) - Posts from user
• scrapeInstagramHashtag(input) - Posts by hashtag
• scrapeInstagramComments(input) - Comments on post

LinkedIn

• scrapeLinkedInProfile(input) - Profile + experience + email
• searchLinkedInJobs(input) - Job listings
• scrapeLinkedInPosts(input) - Posts from profile/company

TikTok

• scrapeTikTokProfile(input) - Profile + videos
• scrapeTikTokHashtag(input) - Videos by hashtag
• scrapeTikTokComments(input) - Comments on video

YouTube

• scrapeYouTubeChannel(input) - Channel + videos
• searchYouTube(input) - Search videos
• scrapeYouTubeComments(input) - Comments on video

Facebook

• scrapeFacebookPosts(input) - Posts from pages
• scrapeFacebookGroups(input) - Group posts
• scrapeFacebookComments(input) - Post comments

Business & Lead Generation

Google Maps

• searchGoogleMaps(input) - Search places (with contact extraction!)
• scrapeGoogleMapsPlace(input) - Single place details
• scrapeGoogleMapsReviews(input) - Place reviews

E-commerce

Amazon

• scrapeAmazonProduct(input) - Product details + reviews
• scrapeAmazonReviews(input) - Product reviews only

Web Scraping

General Web

• scrapeWebsite(input) - Custom multi-page crawling
• scrapePage(url, pageFunction) - Single page extraction

⚙️ Configuration

Environment Variables:

# Required - Get from https://console.apify.com/account/integrations
APIFY_TOKEN=apify_api_xxxxx...

Actor Run Options:

{
  memory: 2048,    // MB: 128, 256, 512, 1024, 2048, 4096, 8192
  timeout: 300,    // seconds
  build: 'latest'  // or specific build number
}

🎯 When to Use This vs MCP

Use File-Based (this skill):

• ✅ Need to filter large datasets (>100 results)
• ✅ Want to transform/aggregate data in code
• ✅ Multiple sequential operations
• ✅ Control flow (loops, conditionals)
• ✅ Maximum token efficiency

Use MCP:

• ❌ Simple single operations with small results (<10 items)
• ❌ One-off exploratory queries
• ❌ Don't want to write code

🔗 Links

• Apify Platform: https://apify.com
• Actor Store: https://apify.com/store
• API Docs: https://docs.apify.com/api/v2

Remember: Filter data in code BEFORE returning to model context. This is where the 99% token savings happen!