internal/agent/bundled_skills/camoufox-cli/SKILL.md
Anti-detect browser automation CLI & Skills for AI agents. Use when the user needs to interact with websites with bot detection, CAPTCHAs, or anti-bot blocks, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task that requires bypassing fingerprint checks.
npx skillsauth add fastclaw-ai/fastclaw camoufox-cliInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
camoufox-cli is built on Camoufox (anti-detect Firefox) with C++-level fingerprint spoofing:
navigator.webdriver = falseUse camoufox-cli whenever you need to drive a real browser — it's the default browser tool in this sandbox.
Every browser automation follows this pattern:
camoufox-cli open <url>camoufox-cli snapshot -i (get element refs like @e1, @e2)camoufox-cli close (close the browser when the entire task is fully complete; keep it open if the user may have follow-up instructions)camoufox-cli open https://example.com/form
camoufox-cli snapshot -i
# Output: - textbox "Email" [ref=e1]
# - textbox "Password" [ref=e2]
# - button "Submit" [ref=e3]
camoufox-cli fill @e1 "[email protected]"
camoufox-cli fill @e2 "password123"
camoufox-cli click @e3
camoufox-cli snapshot -i # Check result
Commands can be chained with && in a single shell invocation. The browser persists between commands via a background daemon, so chaining is safe and more efficient than separate calls.
# Chain open + snapshot in one call
camoufox-cli open https://example.com && camoufox-cli snapshot -i
# Chain multiple interactions
camoufox-cli fill @e1 "[email protected]" && camoufox-cli fill @e2 "password123" && camoufox-cli click @e3
# Navigate and capture
camoufox-cli open https://example.com && camoufox-cli screenshot page.png
When to chain: Use && when you don't need to read the output of an intermediate command before proceeding (e.g., open + screenshot). Run commands separately when you need to parse the output first (e.g., snapshot to discover refs, then interact using those refs).
# Navigation
camoufox-cli open <url> # Navigate to URL (starts daemon if needed)
camoufox-cli back # Go back
camoufox-cli forward # Go forward
camoufox-cli reload # Reload page
camoufox-cli url # Print current URL
camoufox-cli title # Print page title
camoufox-cli close # Close browser and stop daemon
camoufox-cli close --all # Close all sessions
# Snapshot
camoufox-cli snapshot # Full aria tree of page
camoufox-cli snapshot -i # Interactive elements only (recommended)
camoufox-cli snapshot -s "#selector" # Scope to CSS selector
# Interaction (use @refs from snapshot)
camoufox-cli click @e1 # Click element
camoufox-cli fill @e1 "text" # Clear + type into input
camoufox-cli type @e1 "text" # Type without clearing (append)
camoufox-cli select @e1 "option" # Select dropdown option
camoufox-cli check @e1 # Toggle checkbox
camoufox-cli hover @e1 # Hover over element
camoufox-cli press Enter # Press keyboard key
camoufox-cli press "Control+a" # Key combination
# Data Extraction
camoufox-cli text @e1 # Get text content of element
camoufox-cli text body # Get all page text (CSS selector)
camoufox-cli eval "document.title" # Execute JavaScript
# Capture
camoufox-cli screenshot # Screenshot as JSON {"base64": "..."}
camoufox-cli screenshot page.png # Screenshot to file
camoufox-cli screenshot --full p.png # Full page screenshot
camoufox-cli pdf output.pdf # Save page as PDF
# Scroll & Wait
camoufox-cli scroll down # Scroll down 500px
camoufox-cli scroll up # Scroll up 500px
camoufox-cli scroll down 1000 # Scroll down 1000px
camoufox-cli wait @e1 # Wait for element to appear
camoufox-cli wait 2000 # Wait milliseconds
camoufox-cli wait --url "*/dashboard" # Wait for URL pattern
# Tabs
camoufox-cli tabs # List open tabs
camoufox-cli switch 2 # Switch to tab by index
camoufox-cli close-tab # Close current tab
# Cookies & State
camoufox-cli cookies # Dump cookies as JSON
camoufox-cli cookies import file.json # Import cookies
camoufox-cli cookies export file.json # Export cookies
# Sessions
camoufox-cli sessions # List active sessions
camoufox-cli --session work open <url> # Use named session
camoufox-cli close --all # Close all sessions
# Setup
camoufox-cli install # Download Camoufox browser
camoufox-cli install --with-deps # Download browser + system libs (Linux)
camoufox-cli open https://example.com/signup
camoufox-cli snapshot -i
camoufox-cli fill @e1 "Jane Doe"
camoufox-cli fill @e2 "[email protected]"
camoufox-cli select @e3 "California"
camoufox-cli check @e4
camoufox-cli click @e5
camoufox-cli snapshot -i # Verify submission result
camoufox-cli open https://example.com/products
camoufox-cli snapshot -i
camoufox-cli text @e5 # Get specific element text
camoufox-cli eval "document.title" # Get page title via JS
camoufox-cli screenshot results.png # Visual capture
# Login and export cookies
camoufox-cli open https://app.example.com/login
camoufox-cli snapshot -i
camoufox-cli fill @e1 "user"
camoufox-cli fill @e2 "pass"
camoufox-cli click @e3
camoufox-cli cookies export auth.json
# Restore in future session
camoufox-cli open https://app.example.com
camoufox-cli cookies import auth.json
camoufox-cli reload
For long-lived accounts where the site also verifies device stability (not just the cookie), combine this with --persistent so the fingerprint stays fixed alongside the cookies — see the Persistent Identity section below.
camoufox-cli open https://site-a.com
camoufox-cli eval "window.open('https://site-b.com')"
camoufox-cli tabs # List tabs
camoufox-cli switch 1 # Switch to second tab
camoufox-cli snapshot -i
camoufox-cli --session s1 open https://site-a.com
camoufox-cli --session s2 open https://site-b.com
camoufox-cli sessions # List both
camoufox-cli --session s1 snapshot -i
camoufox-cli --session s2 snapshot -i
camoufox-cli --headed open https://example.com
camoufox-cli snapshot -i
camoufox-cli screenshot debug.png
When running multiple agents or automations concurrently, always use named sessions to avoid conflicts:
camoufox-cli --session agent1 open https://site-a.com
camoufox-cli --session agent2 open https://site-b.com
camoufox-cli sessions # Check active sessions
Always close your browser session when done to avoid leaked processes:
camoufox-cli close # Close default session
camoufox-cli --session agent1 close # Close specific session
camoufox-cli close --all # Close all sessions
If a previous session was not closed properly, the daemon may still be running. Use camoufox-cli close to clean it up before starting new work.
Single-page apps render their content via JavaScript — the visible business listing, the address, the phone number, the reviews. The DOM text node body is the application bundle, not the rendered UI.
Use snapshot (not text body) on these sites. The aria tree exposes the rendered, accessible labels:
camoufox-cli open "https://www.google.com/maps/search/insurance+agent+austin+tx/"
camoufox-cli snapshot -i # Returns ~80-100 lines of structured listings
# heading "<Business name>"
# img "4.6 stars"
# img "13 reviews"
# button "Hours Closed · Opens 8 am Mon"
# link "Open website"
# button "Copy phone number"
# WRONG for SPAs — returns the JS bundle, not the rendered content
camoufox-cli text body # → 5KB+ of "function(){window.tactilecsi…"
If snapshot shows the page hasn't finished rendering, wait for a known element (e.g. the listing heading) and re-snapshot.
First call may take 2-3 minutes. Spinning up the headless Firefox daemon — especially through a proxy where GeoIP lookup adds latency — is slow. Once the daemon is up, subsequent calls in the same session are sub-second. Set timeout: 180 (or higher) on the exec call that runs your first camoufox-cli open so the agent doesn't kill the daemon mid-bring-up:
// First camoufox-cli call in a fresh sandbox
{"command": "camoufox-cli open https://example.com && camoufox-cli snapshot -i", "timeout": 180}
// Subsequent calls — default timeout is plenty
{"command": "camoufox-cli click @e5 && camoufox-cli snapshot -i"}
Some pages take time to fully load, especially those with dynamic content or heavy JavaScript. Use explicit waits before taking a snapshot:
# Wait for a specific element to appear
camoufox-cli wait @e1
camoufox-cli snapshot -i
# Wait for a URL pattern (useful after redirects)
camoufox-cli wait --url "*/dashboard"
camoufox-cli snapshot -i
# Wait a fixed duration as a last resort
camoufox-cli wait 3000
camoufox-cli snapshot -i
When dealing with slow pages, always wait before snapshotting. If you snapshot too early, elements may be missing from the output.
Refs (@e1, @e2, etc.) are temporary identifiers assigned by sequential numbering during each snapshot. They are invalidated when the page changes.
Always re-snapshot after:
# CORRECT: re-snapshot after navigation
camoufox-cli click @e5 # Navigates to new page
camoufox-cli snapshot -i # MUST re-snapshot
camoufox-cli click @e1 # Use new refs
# CORRECT: re-snapshot after dynamic changes
camoufox-cli click @e1 # Opens dropdown
camoufox-cli snapshot -i # See dropdown items
camoufox-cli click @e7 # Select item
# WRONG: using refs without snapshot
camoufox-cli open https://example.com
camoufox-cli click @e1 # Ref doesn't exist yet!
# WRONG: using old refs after navigation
camoufox-cli click @e5 # Navigates away
camoufox-cli click @e3 # STALE REF - wrong element!
Always take a fresh snapshot before interacting with elements after navigation or page changes.
The ref was invalidated. Re-snapshot to get fresh refs:
camoufox-cli snapshot -i
# Scroll down to reveal element
camoufox-cli scroll down 1000
camoufox-cli snapshot -i
# Or wait for dynamic content
camoufox-cli wait 2000
camoufox-cli snapshot -i
# Scope to a specific container
camoufox-cli snapshot -s "#main-content"
camoufox-cli snapshot -i -s "form.login"
# Wait for URL pattern after redirect
camoufox-cli wait --url "*/dashboard"
camoufox-cli snapshot -i
# Wait a fixed duration as last resort
camoufox-cli wait 3000
camoufox-cli snapshot -i
--session <name> Named session (default: "default")
--headed Show browser window (default: headless)
--timeout <seconds> Daemon idle timeout (default: 1800)
--json Output as JSON instead of human-readable
--persistent [path] Persistent identity — reuse the same fingerprint + cookies across launches
(default path: ~/.camoufox-cli/profiles/<session>)
--proxy <url> Proxy server (http:// or https://; auth: http://user:pass@host:port)
--no-geoip Disable automatic GeoIP spoofing (auto-enabled with --proxy)
--locale <tag> Force browser locale (e.g. "en-US" or "en-US,zh-CN")
By default, every launch gets a fresh random fingerprint. Add --persistent [path] to reuse the same fingerprint + cookies across launches — fingerprint/OS/canvas+font seeds are frozen on first launch (delete the directory to reset); --locale and proxy-derived timezone/geolocation are stored but refreshed whenever you pass the flag; --proxy / --no-geoip are never stored, so pass them every launch.
Use it when the same device should see the same fingerprint across visits (account-bound tasks, parallel independent identities, or when cookies import/export alone isn't enough because the site also checks device stability). Skip it for one-off scraping or quick debugging.
# Parallel identities, each with its own fingerprint + cookies
camoufox-cli --session a --persistent ~/.camoufox-cli/profiles/alice open https://app.example.com
camoufox-cli --session b --persistent ~/.camoufox-cli/profiles/bob open https://app.example.com
# Reset an identity: just remove the directory
rm -rf ~/.camoufox-cli/profiles/alice
tools
Run this BEFORE any package install (pip / npm / apt / brew / cargo / gem / go install) you would otherwise execute via the exec tool — including when the user asks for a deliverable that needs third-party libraries (PDF generation, image processing, scrapers, dashboards, charts, OCR, audio/video, …). It searches the open agent-skill ecosystem (skills.sh) for an installable skill that already does the job; only fall back to direct package install if find-skills returns no credible match. Also covers explicit "find a skill for X" / "how do I do X" / "is there a skill that can..." requests, and any time the user expresses interest in extending agent capabilities.
tools
Run this BEFORE any package install (pip / npm / apt / brew / cargo / gem / go install) you would otherwise execute via the exec tool — including when the user asks for a deliverable that needs third-party libraries (PDF generation, image processing, scrapers, dashboards, charts, OCR, audio/video, …). It searches the open agent-skill ecosystem (skills.sh) for an installable skill that already does the job; only fall back to direct package install if find-skills returns no credible match. Also covers explicit "find a skill for X" / "how do I do X" / "is there a skill that can..." requests, and any time the user expresses interest in extending agent capabilities.
testing
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
testing
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.