NAME
spurlos
You give it an address, and spurlos opens the page like someone who comes for the first time and has not clicked anything yet. You then see which third-party servers the page asks, which cookies it already sets, and whether HTTPS, security headers and forms are right. I wrote spurlos and published it as open source under the MIT license. The name is German for "without a trace".
SYNOPSIS
$ npx spurlos example-company.com --lang en $ docker run --rm ghcr.io/mmohrx/spurlos example-company.com --lang en $ spurlos <url> … [--lang de|en] [--json | --html | --pdf | --sarif] [-o file] [--baseline file] [--mobile] [--gpc] $ spurlos mcp
CHECKS
- Third-party servers
- Every request to another domain, such as Google Fonts, Tag Manager or YouTube. Each of these servers learns the visitor’s IP address. Also which cookie banner the page loads, whether trackers start before it anyway, and whether Google Consent Mode reports consent up front.
- Cookies and storage
- What sits in the browser before any consent, split into first-party, third-party and known trackers, with how long it lasts.
- Data passed on
- Whether email addresses or values from the page address go to third-party servers, and whether ad servers swap their IDs for the same visitor.
- Fingerprinting
- Whether scripts measure the browser through canvas, graphics card, audio or WebRTC to recognize it without a cookie.
- Browser access
- Whether the page asks for location, camera, microphone or notifications right on load.
- Legal notice and privacy policy
- Whether the page links to both and whether they open without an error.
- Privacy policy text
- Whether its text names the services that load before the first click.
- HTTPS
- Certificate, redirect from http to https, and HSTS.
- Security headers
- Content-Security-Policy, X-Frame-Options, X-Content-Type-Options, Referrer-Policy and Permissions-Policy, and whether the CSP lets scripts through anyway.
- Embedded files
- Whether scripts, stylesheets or images come unencrypted over http, and whether the page embeds third-party scripts without an integrity attribute.
- Forms
- Whether they send encrypted to the site itself, to a third-party server, or unencrypted.
- Software details
- Whether headers or the page name the software with its version, such as PHP/8.1.2 or WordPress 6.2.
- Search engines
- Whether a noindex keeps the page out of search engines, left over from before going live, say.
- Mobile
- Whether the page fits a phone without scrolling sideways. With `--mobile`, spurlos opens it as a phone.
- Accessibility
- A few basics: the page language, images without alt text, form fields without a label, blocked zoom.
EXAMPLE
This is the report for a test page that gets a lot wrong on purpose. Above it you see how much the page loads and how long that takes. Problems come first, and every finding has its own light. With --json you get the same as JSON, and a CI can tell from the exit code whether anything is red. The report also comes as an HTML page or PDF, for passing on. In a CI, --baseline reports only what is new, through a ready-made GitHub Action if you like. It can also hand the findings to GitHub code scanning. With spurlos mcp, an AI assistant such as Claude checks sites itself.
spurlos 0.5.0 https://shop.example/tracked.html 15 requests on load, 4 kB; 9 of them with 1 kB to other domains 7 PROBLEMS 6 NOTES 2 FINE ✖ Third-party servers 9 ├─ ● app.usercentrics.eu – Usercentrics (1 request) ├─ ● cdn.jsdelivr.net – jsDelivr (1 request) ├─ ● cdn.unbekannt.example – 1 request ├─ ● connect.facebook.net – Meta Pixel (1 request, only after the first move) ├─ ● fonts.googleapis.com – Google Fonts (1 request) ├─ ● fp.unbekannt.example – 1 request ├─ ● www.google-analytics.com – Google Analytics (1 request) ├─ ● www.googletagmanager.com – Google Tag Manager (1 request) ├─ ● www.youtube.com – YouTube (1 request) └─ ● Google Consent Mode (gcs=G111) – tells Google consent was given for ads and analytics before anyone chose anything → Each of them learns the visitor's IP address on opening the page, before any click. → “Only after the first move” came when spurlos moved the mouse and scrolled, without clicking. Optimizers such as WP Rocket hold scripts back until then; every real visitor sets them off, long before choosing anything in the cookie banner. → Cookie banner found: Usercentrics. → The trackers above still load before anyone chooses anything in the banner. → Google Consent Mode tells Google's tags whether the visitor agreed. Set to “granted” up front, they set cookies and use the data as if the visitor had agreed. The usual default is “denied” until the cookie banner reports the choice. → Fonts can be served from the own server, then no request goes out for them. → Libraries from a CDN can be served from the own server as well. → The site needs the cookie banner before anyone can agree. There are also banners to host yourself, then no request goes out for it. → Videos, maps, chats and social media posts can be embedded so that they only load after a click. → Trackers can be embedded so that they only load once someone agrees in the cookie banner. ✖ Cookies and storage 4 cookies, 1 entry in browser storage ├─ ● Cookie session (shop.example) – first-party, unknown, until the session ends, without Secure ├─ ● Cookie _ga (shop.example) – first-party, Google Analytics, 400 days (longer than 13 months), without Secure ├─ ● Cookie prefs (shop.example) – first-party, unknown, 400 days (longer than 13 months), SameSite=None ├─ ● Cookie _fbp (shop.example) – first-party, Meta Pixel, 90 days, only after the first move └─ ● localStorage: theme – unknown → Set on the first visit, before any click. → Secure: the cookie also travels over unencrypted http. HttpOnly: scripts on the page can read the session. → Hardly any cookie needs to last longer than 13 months; Chrome cuts every lifetime to 400 days anyway. For trackers it weighs more, as they recognize the visitor over a long time that way. → SameSite=None: the browser sends the cookie along even when other sites embed the site or send requests to it. If it does not need that, SameSite=Lax is the better choice. → “Only after the first move” came when spurlos moved the mouse and scrolled, without clicking. Optimizers such as WP Rocket hold scripts back until then; every real visitor sets them off, long before choosing anything in the cookie banner. ✖ Fingerprinting 3 ├─ ● https://fp.unbekannt.example/fp.js – reads back a canvas with text, third-party ├─ ● https://fp.unbekannt.example/fp.js – reads the graphics card via WebGL, third-party └─ ● https://fp.unbekannt.example/fp.js – renders audio in an OfflineAudioContext, third-party → Such values make up a trait that recognizes the browser without any cookie, before anyone clicks anything. ✖ Legal notice and privacy policy incomplete ├─ ● Legal notice (Impressum): “Impressum” → https://shop.example/rechtliches – answers with status 404 └─ ● Privacy policy: no link found on the page ✖ HTTPS with problems ├─ ● http:// serves the page without redirecting to https://. └─ ● No HSTS header (Strict-Transport-Security): browsers do not remember to use https only. – e.g. Strict-Transport-Security: max-age=31536000 ✖ Embedded files 7 ├─ ● script https://www.googletagmanager.com/gtag/js?… – from a third-party server, without integrity ├─ ● script https://app.usercentrics.eu/browser-ui/latest/loader.js – from a third-party server, without integrity ├─ ● script http://shop.example/alt.js – over http on an https page ├─ ● script https://cdn.jsdelivr.net/npm/slider.js – from a third-party server, without integrity ├─ ● script https://fp.unbekannt.example/fp.js – from a third-party server, without integrity ├─ ● script https://connect.facebook.net/en_US/fbevents.js – from a third-party server, without integrity └─ ● stylesheet https://fonts.googleapis.com/css2?… – from a third-party server, without integrity → On an https page the browser blocks scripts, stylesheets and frames over http, and loads images and videos over https or not at all. → Without an integrity attribute the third-party server can swap the file at any time, and the page runs whatever it sends. For files the provider keeps changing, such as a tag manager, that is not possible; then the only way is to serve them yourself. ✖ Forms 2 ├─ ● POST http://shop.example/newsletter – sends unencrypted over http └─ ● POST https://formspree.io/f/abc – sends to a third-party server ▲ Browser access 2 ├─ ● https://shop.example/tracked.html – asks for the location, first-party ├─ ● https://shop.example/tracked.html – asks whether it may send notifications, first-party └─ ● https://shop.example/tracked.html – registers a service worker, first-party → Asked right on load, before anyone clicks anything: the visitor does not know what the site wants it for. Firefox does not even show the notification prompt without a click. → A service worker stays in the browser and can run after the visit too, for push messages, say, or so the site works offline. ▲ Security headers 5 of 5 missing ├─ ● Content-Security-Policy missing – try it as Content-Security-Policy-Report-Only first ├─ ● X-Frame-Options missing – e.g. SAMEORIGIN, or frame-ancestors in the CSP ├─ ● X-Content-Type-Options missing – value: nosniff ├─ ● Referrer-Policy missing – e.g. strict-origin-when-cross-origin └─ ● Permissions-Policy missing – e.g. camera=(), microphone=(), geolocation=() ▲ Software details 3 entries with a version number ├─ ● Header server: Apache/2.4.41 (Ubuntu) ├─ ● Header x-powered-by: PHP/8.1.2 └─ ● <meta name="generator">: WordPress 6.2 → A version number allows a targeted search for known holes. → Version numbers can usually be switched off, e.g. with expose_php = Off in PHP or server_tokens off in nginx. ▲ Search engines are told not to list the page (noindex) └─ ● <meta name="robots">: noindex, nofollow → With noindex, search engines leave the page out. Wanted on a staging site, on a public one usually a leftover from before it went live. ▲ Mobile 1 note └─ ● The page is 1608 pixels wide, the window 1280: visitors have to scroll sideways → Usually an image, a table or an element with a fixed width sticks out over the edge. → For phones, spurlos checks this with --mobile; some sites also load other trackers and cookie banners there. ▲ Accessibility 3 basics missing ├─ ● 1 image without alt text – https://shop.example/team.jpg ├─ ● 2 form fields without a label – input mail, input mail └─ ● The viewport keeps people from zooming in – width=device-width, initial-scale=1, user-scalable=no → Without alt, screen readers read out the file name. Images that only decorate get alt="". → Without <label>, aria-label or title, screen readers do not say what goes into a field; a placeholder is no label. → With user-scalable=no or a maximum-scale under 2, the page cannot be zoomed to twice its size on a phone. → No accessibility statement linked. → These are only basics that can be checked without a click. They do not say whether the site is accessible; that takes tools such as axe and people using it with a screen reader and keyboard. ✔ Data passed on none ✔ Privacy policy text not read Technical pre-check – not legal advice, and no statement on whether the site complies with the GDPR.
LIMITS
- spurlos only checks passively. It evaluates what a normal visit returns anyway and never requests addresses like
/.envor/wp-admin. - It does not click the cookie banner. spurlos sees the page as it is before the visitor decides anything.
- It does not read the legal notice, it only checks that it is there. In the privacy policy it only looks for the names of the services found, and does not judge whether they are described correctly.
- It is not legal advice. A green report does not mean a site complies with the GDPR.
STACK
typescript · node · playwright · chromium · docker · mit