MCP HubMCP Hub
Retour aux compétences

headless-web-scraping

pjt222
Mis à jour 2 days ago
3 vues
17
2
17
Voir sur GitHub
Designapiautomationdesigndata

À propos

Cette compétence permet un web scraping robuste grâce à la bibliothèque Python scrapling, en choisissant automatiquement entre HTTP, Chromium furtif ou l'automatisation complète d'un navigateur selon les défenses du site. Elle gère les pages rendues par JavaScript, les protections anti-bots et la navigation complexe dans le DOM pour l'extraction de données structurées. Utilisez-la lorsque WebFetch échoue en raison de contenu dynamique ou de mécanismes de blocage avancés.

Installation rapide

Claude Code

Recommandé
Principal
npx skills add pjt222/agent-almanac -a claude-code
Commande PluginAlternatif
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternatif
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/headless-web-scraping

Copiez et collez cette commande dans Claude Code pour installer cette compétence

Documentation

Headless Web Scraping

Extract data from web pages that resist simple HTTP requests — JS-rendered content, Cloudflare-protected sites, and dynamic SPAs — using scrapling's three-tier fetcher architecture and CSS-based data extraction.

When to Use

  • Target page requires JavaScript rendering (SPA, React, Vue)
  • Site has anti-bot protections (Cloudflare Turnstile, TLS fingerprinting)
  • You need structured extraction of multiple elements via CSS selectors
  • Simple WebFetch or requests.get() returns empty or blocked responses
  • Extracting tabular data, link lists, or repeated DOM structures at scale

Inputs

  • Required: Target URL or list of URLs to scrape
  • Required: Data to extract (CSS selectors, field names, or description of target elements)
  • Optional: Fetcher tier override (default: auto-select based on site behavior)
  • Optional: Output format (default: JSON; alternatives: CSV, Python dict)
  • Optional: Rate limit delay in seconds (default: 1)

Procedure

Step 1: Select Fetcher Tier

Determine which scrapling fetcher matches the target site's defenses.

# Decision matrix:
# 1. Fetcher        — static HTML, no JS, no anti-bot (fastest)
# 2. StealthyFetcher — Cloudflare/Turnstile, TLS fingerprint checks
# 3. DynamicFetcher  — JS-rendered SPAs, click/scroll interactions

# Quick probe: try Fetcher first, escalate on failure
from scrapling import Fetcher

fetcher = Fetcher()
response = fetcher.get("https://example.com/target-page")

if response.status == 200 and response.get_all_text():
    print("Fetcher tier sufficient")
else:
    print("Escalate to StealthyFetcher or DynamicFetcher")
SignalRecommended Tier
Static HTML, no protectionFetcher
403/503, Cloudflare challenge pageStealthyFetcher
Page loads but content area is emptyDynamicFetcher
Need to click buttons or scrollDynamicFetcher
altcha CAPTCHA presentNone (cannot be automated)

Got: One of the three tiers is identified. For most modern sites, StealthyFetcher is the correct starting point.

If fail: If all three tiers return blocked responses, check whether the site uses altcha CAPTCHA (proof-of-work challenge that cannot be bypassed). If so, document the limitation and provide manual extraction instructions instead.

Step 2: Configure the Fetcher

Set up the selected fetcher with appropriate options.

from scrapling import Fetcher, StealthyFetcher, DynamicFetcher

# Tier 1: Fast HTTP with TLS fingerprint impersonation
fetcher = Fetcher()
fetcher.configure(
    timeout=30,
    retries=3,
    follow_redirects=True
)

# Tier 2: Headless Chromium with anti-detection
fetcher = StealthyFetcher()
fetcher.configure(
    headless=True,
    timeout=60,
    network_idle=True  # wait for all network requests to settle
)

# Tier 3: Full browser automation
fetcher = DynamicFetcher()
fetcher.configure(
    headless=True,
    timeout=90,
    network_idle=True,
    wait_selector="div.results"  # wait for specific element before extracting
)

Got: Fetcher instance is configured and ready. No errors on instantiation. For StealthyFetcher and DynamicFetcher, a Chromium binary is available (scrapling manages this automatically on first run).

If fail:

  • playwright or browser binary not found -- run python -m playwright install chromium
  • Timeout on configure() -- increase timeout value or check network connectivity
  • Import error -- install scrapling: pip install scrapling

Step 3: Fetch and Extract Data

Navigate to the target URL and extract structured data using CSS selectors.

# Fetch the page
response = fetcher.get("https://example.com/target-page")

# Single element extraction
title = response.find("h1.page-title")
if title:
    print(title.get_all_text())

# Multiple elements
items = response.find_all("div.result-item")
for item in items:
    name = item.find("span.name")
    price = item.find("span.price")
    print(f"{name.get_all_text()}: {price.get_all_text()}")

# Get attribute values
links = response.find_all("a.product-link")
urls = [link.get("href") for link in links]

# Get raw HTML content of an element
detail_html = response.find("div.description").html_content

Key API reference:

MethodPurpose
response.find("selector")First matching element
response.find_all("selector")All matching elements
element.get("attr")Attribute value (href, src, data-*)
element.get_all_text()All text content, recursively
element.html_contentRaw inner HTML

Got: Extracted data matches the visible page content. Elements are non-None and text content is non-empty for populated pages.

If fail:

  • find() returns None -- inspect the actual HTML (response.html_content) to verify the selector; the page may use different class names than expected
  • Empty text from get_all_text() -- content may be inside shadow DOM or an iframe; try DynamicFetcher with a wait_selector
  • Do NOT use .css_first() -- this is not part of the scrapling API (common confusion with other libraries)

Step 4: Handle Failures and Edge Cases

Implement fallback logic for CAPTCHA detection, empty responses, and session requirements.

import time

def scrape_with_fallback(url, selector):
    """Try each fetcher tier in order, with CAPTCHA detection."""
    tiers = [
        ("Fetcher", Fetcher),
        ("StealthyFetcher", StealthyFetcher),
        ("DynamicFetcher", DynamicFetcher),
    ]

    for tier_name, tier_class in tiers:
        fetcher = tier_class()
        fetcher.configure(headless=True, timeout=60)

        try:
            response = fetcher.get(url)
        except Exception as error:
            print(f"{tier_name} failed: {error}")
            continue

        # Detect CAPTCHA / challenge pages
        page_text = response.get_all_text().lower()
        if "altcha" in page_text or "proof of work" in page_text:
            print(f"altcha CAPTCHA detected -- cannot automate")
            return None

        if response.status == 403 or response.status == 503:
            print(f"{tier_name} blocked (HTTP {response.status}), escalating")
            continue

        result = response.find(selector)
        if result and result.get_all_text().strip():
            return result.get_all_text()

        print(f"{tier_name} returned empty content, escalating")

    print("All tiers exhausted. Manual extraction required.")
    return None

Got: Function returns extracted text on success, or None with a diagnostic message when all tiers fail. CAPTCHA pages are detected and reported rather than retried indefinitely.

If fail:

  • All tiers return 403 -- the site blocks all automated access (common with WIPO, TMview, some government databases); document the URL as requiring manual access
  • Timeout errors -- the page may be behind a slow CDN; increase timeout to 120s
  • Session/cookie errors -- the site may require login; add cookie handling or authenticate first

Step 5: Rate Limiting and Ethical Scraping

Implement delays and respect site policies before running at scale.

import time
import urllib.robotparser

def check_robots_txt(base_url, target_path):
    """Check if scraping is allowed by robots.txt."""
    rp = urllib.robotparser.RobotFileParser()
    rp.set_url(f"{base_url}/robots.txt")
    rp.read()
    return rp.can_fetch("*", f"{base_url}{target_path}")

def scrape_urls(urls, selector, delay=1.0):
    """Scrape multiple URLs with rate limiting."""
    results = []
    fetcher = StealthyFetcher()
    fetcher.configure(headless=True, timeout=60)

    for url in urls:
        response = fetcher.get(url)
        data = response.find(selector)
        if data:
            results.append(data.get_all_text())

        time.sleep(delay)  # respect the server

    return results

Ethical scraping checklist:

  1. Check robots.txt before scraping -- respect Disallow directives
  2. Use a minimum 1-second delay between requests
  3. Identify your scraper with a descriptive User-Agent when possible
  4. Do not scrape personal data without legal basis
  5. Cache responses locally to avoid redundant requests
  6. Stop immediately if you receive a 429 (Too Many Requests)

Got: Scraping runs at a controlled rate. robots.txt is checked before bulk operations. No 429 responses are triggered.

If fail:

  • 429 Too Many Requests -- increase delay to 3-5 seconds, or stop and retry later
  • robots.txt disallows the path -- respect the directive; do not override it
  • IP ban -- stop scraping immediately; the rate limiting was insufficient. If access is legitimate (public data, ToS-permitted, robots.txt-respected) and you must continue, see rotate-scraping-proxies for network-layer escalation

Validation

  • Correct fetcher tier is selected (not over- or under-powered for the target)
  • configure() method is used (not deprecated constructor kwargs)
  • CSS selectors match actual page structure (verified against page source)
  • .find() / .find_all() API is used (not .css_first() or other library methods)
  • CAPTCHA detection is in place (altcha pages are reported, not retried)
  • Rate limiting is implemented for multi-URL scraping
  • robots.txt is checked before bulk operations
  • Extracted data is non-empty and structurally correct

Pitfalls

  • Using .css_first() instead of .find(): scrapling uses .find() and .find_all() for element selection -- .css_first() belongs to a different library and will raise AttributeError
  • Starting with DynamicFetcher: Try Fetcher first, then escalate -- DynamicFetcher is 10-50x slower due to full browser startup
  • Constructor kwargs instead of configure(): scrapling v0.4.x deprecated passing options to the constructor; use the configure() method
  • Ignoring altcha CAPTCHA: No fetcher tier can solve altcha proof-of-work challenges -- detect them early and fall back to manual instructions
  • No rate limiting: Even if the site does not return 429, aggressive scraping can get your IP banned or cause service degradation
  • Assuming stable selectors: Website CSS classes change frequently -- validate selectors against current page source before each scraping campaign

Related Skills

  • rotate-scraping-proxies -- network-layer escalation when client-side stealth is exhausted and IP bans block legitimate, ToS-permitted access
  • use-graphql-api -- structured API queries when the site offers a GraphQL endpoint (preferred over scraping)
  • serialize-data-formats -- converting extracted data to JSON, CSV, or other formats
  • deploy-searxng -- self-hosted search engine that aggregates results from multiple sources
  • forage-solutions -- broader pattern for gathering information from diverse sources
<!-- Keep under 500 lines. Extract large examples to references/EXAMPLES.md if needed. -->

Dépôt GitHub

pjt222/agent-almanac
Chemin: i18n/caveman-lite/skills/headless-web-scraping
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Compétences associées

executing-plans

Design

Utilisez la compétence executing-plans lorsque vous disposez d'un plan de mise en œuvre complet à exécuter par lots contrôlés avec des points de contrôle de revue. Elle charge et examine le plan de manière critique, puis exécute les tâches par petits lots (3 tâches par défaut) tout en rapportant la progression entre chaque lot pour une revue par l'architecte. Cela garantit une mise en œuvre systématique avec des points de contrôle de qualité intégrés.

Voir la compétence

requesting-code-review

Design

Cette compétence délègue un sous-agent réviseur de code pour analyser les modifications apportées au code par rapport aux exigences avant de poursuivre. Elle doit être utilisée après avoir terminé des tâches, implémenté des fonctionnalités majeures, ou avant une fusion vers la branche principale. La revue aide à détecter précocement les problèmes en comparant l'implémentation actuelle avec le plan initial.

Voir la compétence

connect-mcp-server

Design

Cette compétence fournit un guide complet permettant aux développeurs de connecter des serveurs MCP à Claude Code via les transports HTTP, stdio ou SSE. Elle couvre l'installation, la configuration, l'authentification et la sécurité pour intégrer des services externes tels que GitHub, Notion et des API personnalisées. Utilisez-la lors de la configuration d'intégrations MCP, de la configuration d'outils externes ou du travail avec le Protocole de Contexte de Modèle de Claude.

Voir la compétence

web-cli-teleport

Design

Cette compétence aide les développeurs à choisir entre les interfaces Web et CLI de Claude Code en fonction de l'analyse des tâches, puis permet une téléportation transparente des sessions entre ces environnements. Elle optimise le flux de travail en gérant l'état et le contexte de la session lors du passage entre le web, la CLI ou le mobile. Utilisez-la pour des projets complexes nécessitant différents outils à diverses étapes.

Voir la compétence