Skip to content

SKILL.md

Latest commit

 

History

History
75 lines (67 loc) · 4.35 KB

SKILL.md

File metadata and controls

75 lines (67 loc) · 4.35 KB
Raw
name fortress-stealth-browser
description Use when a web fetch is blocked — Cloudflare, DataDome, PerimeterX, Akamai, a 403/429, a CAPTCHA/"Press & Hold", "Access denied", an empty JavaScript shell, or a page whose data only appears after client-side rendering. Drives a real recompiled-Chromium stealth engine (Fortress) via the Fortress MCP to fetch, extract, crawl, recon, search, and automate. Also use to get clean markdown/tables/JSON from any page, or a CDP endpoint to run existing Playwright/Puppeteer/browser-use code through the stealth browser.

Fortress stealth-browser skill

When to use this

Reach for the fortress MCP tools the moment a normal fetch fails or returns junk:

  • HTTP 403 / 429 / 503, a Cloudflare "Just a moment", DataDome/PerimeterX walls, "Access to this page has been denied", "Press & Hold to confirm you are human".
  • The page is a JavaScript SPA and the plain HTML is an empty shell.
  • You need structured data (markdown, tables, or a schema-shaped record), a whole-site crawl, or the site's private JSON API.
  • You have Playwright/Puppeteer/browser-use code that keeps getting detected — get a stealth CDP endpoint and point your code at it.

Setup

pip install "tilion[mcp]"

Add to the client MCP config: { "mcpServers": { "fortress": { "command": "tilion-mcp" } } }

Tool cheat-sheet (pick by intent)

  • Blocked / need the page → fetch_protected_page(url) → {status, http_status, blocked, waf, title, text}.
  • Clean content → read_page(url) (markdown) or extract_page(url, schema?) (record).
  • A file (PDF/DOCX/XLSX/CSV) → extract_document(source).
  • Whole site → crawl_site(url, depth, max_pages).
  • Find the data source → recon_site_apis(url) → discovered XHR/JSON endpoints.
  • Search the web → search_web(query, count) (real browser, no SERP key).
  • Interact → page_elements → click_button / fill_field / press_key / wait_for / evaluate_js; check with current_page.
  • Multi-step flow (login, paginate, infinite-scroll, checkout) → list_browser_tasks() then run_browser_task(task, url, params?).
  • Identify the anti-bot → detect_waf(url) → {vendor, challenged, strategy}.
  • Solve a captcha → solve_captcha(url) (needs CAPTCHA_API_KEY).
  • Check egress → get_egress_info() → the real public IP the target sees.
  • Auth reuse → save_profile(name) / load_profile(name).
  • Capture → screenshot_page / save_page(format=pdf|html|text|png) / download_file.
  • Bring your own automation → get_stealth_cdp_endpoint() → a CDP url for Playwright / Puppeteer / browser-use / Crawl4AI.

Handling a block (the golden path)

fetch_protected_page returns a waf field — read it, don't guess:

  • status:"ok" → you have the page; extract/read/act.
  • status:"empty" → JS withheld content; retry or use run_browser_task to interact.
  • status:"blocked" → check waf.vendor: cloudflare/incapsula usually auto-clears (retry once); datadome/perimeterx are behavioral (fetch already nudged — if still blocked you need residential egress); akamai/kasada are IP-gated → set TILION_PROXY and retry; a captcha present → solve_captcha (or fetch auto-solves with CAPTCHA_API_KEY).

Configure the hard-target levers (env)

  • TILION_PROXY=http://user:pass@host:port — residential/mobile egress (the #1 unblock lever; verify with get_egress_info). TILION_REGION=us aligns timezone/locale to the IP.
  • CAPTCHA_API_KEY=... (+ CAPTCHA_PROVIDER=2captcha|anticaptcha|capsolver) — auto-solve.

Good habits

  • Prefer extract_page/read_page over dumping raw HTML — outputs are capped and clean.
  • For an authenticated site, load_profile before fetching so cookies + solved challenges are reused.
  • Reads are safe/auto-approved; run_browser_task, save_page, download_file write or act — confirm intent first.
  • A great fingerprint on a datacenter IP can still be blocked by the hardest sites; on a residential IP the local engine clears most walls. Hosted residential egress is coming soon.

Reference

  • mcp/USAGE.md — the full 29-tool guide: per-tool detail, workflow recipes, result-reading, and config.
  • mcp/README.md — install, tool table, benchmarks.