Public homepage at https://sigilvault.xyz/. The page describes “HSM-grade custody, shipped like an API.”
The page has real, crawlable copy (509 words). Web Scan’s AI-crawler view is “partial”: it can tell what the product is and that a visitor can sign up, but not what it costs. The failures are mostly in search snippets, link previews, crawl basics and one accessibility miss, plus a missing Content-Security-Policy header. Nothing in this scan reported leaked secret keys.
| Area | Score | Pass | Warn | Fail |
|---|---|---|---|---|
| Search basics | 68 | 6 | 1 | 1 |
| Content | 84 | 5 | 2 | 0 |
| Google extras | 35 | 0 | 2 | 0 |
| Sharing on social | 23 | 1 | 5 | 2 |
| Findable by Google | 78 | 4 | 3 | 0 |
| AI assistants | 100 | 2 | 0 | 0 |
| Images | 100 | 0 | 0 | 0 |
| Security | 94 | 7 | 1 | 0 |
| Leaked secrets | 100 | 2 | 0 | 0 |
| Speed | 84 | 8 | 1 | 0 |
| Accessibility basics | 67 | 3 | 1 | 1 |
Web Scan caps a grade when a critical check fails, so strong security and content scores cannot pull this page up to an A while Open Graph and the meta description are missing.
Web Scan fetched the HTML the way GPTBot, ClaudeBot and PerplexityBot would (no JavaScript). This page is not a blank client-rendered shell.
Ranked by likely business impact for a custody and signing-infrastructure startup that is onboarding design partners, not by raw check weight.
Finding. There is no og:image or twitter:image. The share preview has no image.
Business impact. This is the page you will paste into X, Telegram, Farcaster, Slack, LinkedIn and investor chats. Without an image, the unfurl is a title and a URL. For a product that has to look institutional, that is a weak first impression. Web Scan’s own guidance is that a preview image is the highest-leverage tag on the page.
Fix. Add a 1200×630 image at an absolute HTTPS URL that does not require cookies:
<meta property="og:image" content="https://sigilvault.xyz/og.png"> <meta property="og:image:width" content="1200"> <meta property="og:image:height" content="630">
Then fetch that image URL in a private window (or re-run Web Scan) to confirm it does not return 403 to crawlers.
Finding. og:title is missing (fail). Also missing: og:description, og:url, og:type, og:site_name, and twitter:card.
Business impact. Slack, LinkedIn, Discord, WhatsApp and iMessage all read Open Graph. Without these tags a link often unfurls as a bare URL, and X may not use a large image card even after you add an image.
Fix. Suggested values, taken from the existing title and H1 — edit the wording if you prefer:
<meta property="og:title" content="Sigil — Signing infrastructure for modern custody"> <meta property="og:description" content="HSM-grade custody, shipped like an API. Keys stay in hardware; your app calls one endpoint."> <meta property="og:url" content="https://sigilvault.xyz/"> <meta property="og:type" content="website"> <meta property="og:site_name" content="Sigil"> <meta name="twitter:card" content="summary_large_image">
Finding. No <meta name="description">. The scan’s preview description is empty.
Business impact. Search engines will invent a snippet, or show nothing useful, under your otherwise good 49-character title. That tag is the sentence that decides a click.
Fix. Keep it roughly 120–158 characters:
<meta name="description" content="Sigil is HSM-grade signing infrastructure with a cloud control plane. Keys stay in hardware; your app calls one API. Built for exchanges, custodians and wallet teams.">
/robots.txt is a web page, not a robots file (warning)Finding. /robots.txt returned 200 with a web page, not a robots.txt file. /sitemap.xml similarly answered with something that is not a sitemap.
Business impact. Crawlers read the body. An app-shell or marketing HTML at /robots.txt is treated as a missing or nonsense robots file, and often means unknown paths are also serving the homepage (see issue 5).
Fix. Serve a real text/plain robots.txt. For a one-page site, at minimum:
User-agent: * Allow: / Sitemap: https://sigilvault.xyz/sitemap.xml
Do not let a catch-all route handle /robots.txt or /sitemap.xml. A sitemap is optional on a true one-page site; if you add one, it must be XML.
Finding. A made-up path on this domain answered 200 OK instead of 404.
Business impact. Search engines can index junk URLs under your domain. For a custody brand, random paths that look like real pages are sloppy and can be used in confusing links.
Fix. Make addresses that do not exist return a real 404 status and a simple “not found” page.
Finding. No JSON-LD structured data, and no Organization entity block.
Business impact. Google’s extras and many AI answers credit a known brand, not a bare domain. The visible copy already names Sigil Vault Ltd.; the hidden markup is how machines attach a name, logo and official profiles to that copy.
Fix. Only include profiles and a logo URL you actually own:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Sigil",
"url": "https://sigilvault.xyz/",
"legalName": "Sigil Vault Ltd.",
"sameAs": ["https://github.com/SigilVault"]
}
</script>
Finding. No Content-Security-Policy header. Other transport basics are in good shape: HTTPS, HTTP→HTTPS redirect, HSTS (max-age=31536000), nosniff, framing restricted, referrer policy strict-origin-when-cross-origin, no mixed content, no leaked provider keys in the HTML.
Business impact. CSP is the main browser defence against an injected script. On a custody product, visitors and design partners will look at headers. This is not a proven breach; it is a missing control.
Fix. Start with Report-Only so nothing breaks while you list your script sources:
Content-Security-Policy-Report-Only: default-src 'self'; script-src 'self'; object-src 'none'; base-uri 'self'
Then switch to an enforcing Content-Security-Policy once the report is clean.
Finding. The document took 1,541 ms to return. HTML weight itself is fine (65 KB, gzip, no render-blocking scripts in <head>).
Business impact. Everything else waits for the first byte. Web Scan’s bar is about 800 ms. A 1.5s wait is noticeable on a mobile connection and is cheap to fix on a mostly static marketing page.
Fix. Cache the HTML on the CDN, or serve the page as a static file.
Finding. 1 of 1 form fields has no <label>. A placeholder is not enough.
Business impact. This is likely the design-partner field — the one conversion on the page. Screen readers may not name it, and some browsers and password managers handle unlabelled fields poorly.
Fix. Give the field a <label for="…">, or an aria-label if there is no room for visible text.
Finding. No internal links to other pages on this domain. No link to an about or privacy page. No <main> landmark. No canonical URL.
Business impact. Visitors, and Google’s quality guidelines, look for who runs a site before they trust a custody product. A canonical tag stops tracking-parameter copies from splitting the page.
Fix. Add a footer with About, Privacy and a contact or email link; wrap the primary content in <main>; add <link rel="canonical" href="https://sigilvault.xyz/">.
og:image, og:title and the other share tags, plus a meta description (issues 1–3). Small HTML change, and the one visitors notice first.robots.txt, and 404s for unknown paths.A coding agent can apply most of this from the snippets above. After deploy, the same checks can be re-run with Web Scan’s /api/verify, or as a 1 USDC written re-check.