$ ~/registry/skill/composiohq-composio-skills-firecrawl-automation

SKILL

Firecrawl Automation

Name: Firecrawl Automation
Author: Composio

Automate web crawling, structured extraction, batch scraping, and site mapping.

Stars

★ 67,763

Source

GitHub

Updated

2026-07-15

// securityCaution

Prompt only, no code
Open source
Community-trusted· 64.7k

Summary

Overall this appears low risk: the material indicates an open-source, highly adopted prompt-only skill with no stated local execution or local file access. The main caution is that its documentation references Composio/Firecrawl for web crawling and account connection, which introduces ordinary third-party network access and data handling.

CredentialsCaution

The metadata says there are no required keys/environment variables, but the README explicitly says to 'Connect your Firecrawl account when prompted,' indicating possible third-party account authorization during use. No direct request for local sensitive credentials is shown; the main concern is standard third-party token authorization and credit abuse exposure.

Network egressCaution

Its stated purpose is web scraping, crawling, and structured extraction, and the README mentions connecting through the Composio MCP endpoint 'https://rube.app/mcp' while accessing user-specified website URLs. This is ordinary egress within the declared functionality, and the material does not show unrelated or covert exfiltration endpoints.

Code executionLow risk

The system flags it as prompt-only, and the material does not describe spawning local processes, running shell/scripts, or requesting expanded system privileges. The documentation only describes web-crawling tool actions, with no clear red flags for local code execution.

Data accessLow risk

There is no stated ability to read or write local files, the clipboard, system directories, or other local data sources; its data scope appears to be user-provided URLs and the fetched web content. Based on the material, there is no local data access beyond the declared function.

Supply chainCaution

There are strong positive signals: it is open-source on GitHub with high community adoption (64.7k stars), which lowers supply-chain risk. However, the license is unspecified, maintenance status is unknown, and the repository appears to be a skills collection rather than the canonical standalone implementation, so some provenance checks are still advisable.

Recommendations

Only connect a Firecrawl account in a trusted environment, and review the authorization scope and spending limits.
Start with small URL sets and low crawl limits to avoid excessive crawling or unexpected credit consumption.
Verify the relationship between the referenced MCP endpoint and the actual publisher, and prefer official documentation and official domains.
If using sensitive URLs or internal sites, confirm the third-party service's data retention, logging, and compliance policies first.

Audited by: gpt-5.4 · 2026-06-16

// install

Copy the install command and let the AI configure it · recommended for beginners

Please install the "Firecrawl Automation" skill from askskill:
1. Download https://raw.githubusercontent.com/ComposioHQ/awesome-claude-skills/master/composio-skills/firecrawl-automation/SKILL.md
2. Save it as ~/.claude/skills/firecrawl-automation/SKILL.md
3. Reload skills and tell me it's ready

// download

Download SKILL.mdInstall manifest ↗

// docs

Firecrawl Automation

Run Firecrawl web crawling and extraction directly from Claude Code. Scrape individual pages, crawl entire sites, extract structured data with AI, batch process URL lists, and map website structures without leaving your terminal.

Toolkit docs: composio.dev/toolkits/firecrawl

Setup

Add the Composio MCP server to your configuration:
```
https://rube.app/mcp
```
Connect your Firecrawl account when prompted. The agent will provide an authentication link.
Be mindful of credit consumption -- scope your crawls tightly and test on small URL sets before scaling.

Core Workflows

1. Scrape a Single Page

Fetch content from a URL in multiple formats with optional browser actions for dynamic pages.

Tool: FIRECRAWL_SCRAPE

Key parameters:

url (required) -- fully qualified URL to scrape
formats -- output formats: markdown (default), html, rawHtml, links, screenshot, json
onlyMainContent (default true) -- extract main content only, excluding nav/footer/ads
waitFor -- milliseconds to wait for JS rendering (default 0)
timeout -- max wait in ms (default 30000)
actions -- browser actions before scraping (click, write, wait, press, scroll)
includeTags / excludeTags -- filter by HTML tags
jsonOptions -- for structured extraction with schema and/or prompt

Example prompt: "Scrape the main content from https://example.com/pricing as markdown"

2. Crawl an Entire Site

Discover and scrape multiple pages from a website with configurable depth, path filters, and concurrency.

Tool: FIRECRAWL_CRAWL_V2

Key parameters:

url (required) -- starting URL for the crawl
limit (default 10) -- max pages to crawl
maxDiscoveryDepth -- depth limit from the root page
includePaths / excludePaths -- regex patterns for URL paths
allowSubdomains -- include subdomains (default false)
crawlEntireDomain -- follow sibling/parent links, not just children (default false)
sitemap -- include (default), skip, or only
prompt -- natural language to auto-configure crawler settings
scrapeOptions_formats -- output format for each page
scrapeOptions_onlyMainContent -- main content extraction per page

Example prompt: "Crawl the docs section of firecrawl.dev, max 50 pages, only paths matching docs"

3. Extract Structured Data

Extract structured JSON data from web pages using AI with a natural language prompt or JSON schema.

Tool: FIRECRAWL_EXTRACT

Key parameters:

urls (required) -- array of URLs to extract from (max 10 in beta). Supports wildcards like https://example.com/blog/*
prompt -- natural language description of what to extract
schema -- JSON Schema defining the desired output structure
enable_web_search -- allow crawling links outside initial domains (default false)

At least one of prompt or schema must be provided.

Check extraction status with FIRECRAWL_EXTRACT_GET using the returned job id.

Example prompt: "Extract company name, pricing tiers, and feature lists from https://example.com/pricing"

4. Batch Scrape Multiple URLs

Scrape many URLs concurrently with shared configuration for efficient bulk data collection.

Tool: FIRECRAWL_BATCH_SCRAPE

Key parameters:

urls (required) -- array of URLs to scrape
formats -- output format for all pages (default markdown)
onlyMainContent (default true) -- main content extraction
maxConcurrency -- parallel scrape limit
ignoreInvalidURLs (default true) -- skip bad URLs instead of failing the batch
location -- geolocation settings with country code
actions -- browser actions applied to each page
blockAds (default true) -- block advertisements