Is Seven Sky's own site ready for AI assistants?

Assistants can read this site, and an assistant that supports MCP can use it as a tool.

Both programs pass most of what they test. A person reading the site found three gaps that neither program looks for. They are fixed in the version of the site that carries this report. Three smaller fixes are still open.

No fix on this list makes an AI product cite, rank or recommend a site. The fixes make the site easier to read and to use. What an assistant does with it is up to its maker.

• The site's own free check, run once on 15 October 2026 at 07:02 UTC: twelve checks, all passed.
• Cloudflare's public scanner, isitagentready.com, run once in the same minute, with no sign-in: 22 checks. 9 passed and 7 found nothing. 6 were not scored for this site: five shopping checks, and one that is for information only. It put the site at level 4 of 0 to 5, which it calls Agent-Integrated.
• A person read the home page, the offers, check, samples, terms and privacy pages, robots.txt, the site map, llms.txt and the discovery files.

• The home page answers over HTTPS
  Status 200, with its title, description, canonical address and language set.
• robots.txt lets assistants in and says what they may do
  Every crawler is allowed. Its Content-Signal line reads search=yes, ai-input=yes, ai-train=yes. It names the site map.
• A site map
  /sitemap.xml listed 13 addresses on the day it was read.
• An llms.txt file
  A plain-text guide to the offers, the products that are not for sale yet and the tools, with links to Markdown versions.
• Markdown on request
  A request for the home page with Accept: text/markdown gets Markdown back. Ten inner pages name a Markdown version in their head.
• The pages say what they are
  JSON-LD: Organization on the home page, OfferCatalog on the offers page.
• Readable without scripts
  About 371 words are in the home page as delivered.
• Link headers
  The home page's response points to its Markdown version, to llms.txt and to the API catalog.
• The check page has no structured data and no Markdown version
  Fixes 4 and 5 below.

Yes, if the assistant supports MCP. An assistant without MCP reads the pages instead.

• An MCP server at /mcp, with no sign-in optional, early
  Its server card, at /.well-known/mcp/server-card.json, lists six tools today: list_options, build_chief_of_staff, check_site, list_offers, list_products and request_offer.
• An API catalog optional, early
  /.well-known/api-catalog lists five addresses: /mcp, /api/check, /api/order, /api/vote and /api/fix-pack/preview.
• An agent-skills index
  It lists one skill, the chief-of-staff builder.
• The offers as data
  /offers/offers.json gives each offer's price and its pay address. A person pays on Stripe's page; the site takes no payment itself.
• No WebMCP tools on the page optional, early
  WebMCP lets a page hand its tools to an assistant in the browser. It is new. Fix 7 below.
• No A2A agent card, ARD manifest or DNS-AID records optional, early
  Cloudflare's scanner looks for them. Fix 8 below.
• No OAuth discovery, protected-resource metadata or auth.md does not apply
  These are for sites where an assistant signs in. Nothing here needs a sign-in.

Optional and early: in April 2026 Cloudflare found MCP server cards and API catalogs on fewer than 15 of the top 200,000 domains (Cloudflare). Treat these files as optional.

First what a person misses at the first look, then what an assistant reads, then what is optional and early. Each fix names the file it touches.

1. Put Offers in the home page's top menu done
   The menu led to the seven layers, how the company works, the record and contact, but not to what can be bought.
   File: /index.html, the menu (<nav>).
2. Add an About page done
   The founder's name, the company and his record were spread over the home, terms and offers pages. No page put them together.
   Files: /about/index.html and /about/index.md, new; the site map and llms.txt, which /_worker.js writes.
3. Show the $450 offer on the check page before a check runs done
   The line that named it sat in the results section, hidden until a check ran.
   File: /check/index.html.
4. Give the check page structured data open
   One JSON-LD block that says it is a free web tool (schema.org WebApplication), as the home and offers pages describe themselves.
   File: /check/index.html, the head.
5. Give the check page a Markdown version open
   Ten other pages have one and name it in their head.
   Files: /check/index.md, new; one alternate link in /check/index.html.
6. Name the home page's Markdown version in its HTML head open
   Today only the HTTP Link header names it. A program that reads only the HTML does not see it.
   File: /index.html: <link rel="alternate" type="text/markdown" href="/index.md">
7. Offer the free check as a WebMCP tool optional, early
   A browser assistant could then run the check from the page. WebMCP is new.
   File: /js/check.js.
8. An A2A agent card, an ARD manifest, DNS-AID records optional, early
   The MCP server card already lists the site's tools, so these can wait.
   Files: /.well-known/agent-card.json and an ARD manifest. DNS-AID records live in the domain's DNS, not in a site file.

In a paid report, each open fix also comes with its file, written for that site, and a 30-minute call walks through the list. This sample has neither.

• Whether any AI product cites, ranks or recommends the site. No questions were put to ChatGPT, Claude, Gemini or Perplexity.
• How the site compares with others that sell the same work.
• Search rankings, traffic and sales.
• Speed, accessibility and security, beyond what is named above.
• The products, fix pack, skill pack, housing, course, vote and builder pages, line by line.
• The order form, the vote and the payment pages. None was used.
• Visits from other networks. Each program read the site once, from its own servers. A site that turns away visitors it does not recognise can read differently from elsewhere.
• llms.txt in Cloudflare's scan. Its run on this day did not report on it; the site's own check did.