TRANSPARENT BY DESIGN
What we check.
What it means.
An automated finding is a starting point for a decision. Here’s how our checks work, where they stop and how we handle your data.
Website checks
We start from the address you enter, follow up to four redirects, then crawl up to 15 public pages on the same host by following ordinary links, at most three links from the entry page. Requests are made one at a time within a 40-second budget. We honour robots.txt, skip links marked nofollow, skip downloads and media, and never follow links that look like sign-out, delete, unsubscribe or checkout actions. Subdomains and other sites are not followed. On each page we check HTTPS certificate validation, selected security headers, cookie attributes, static references to insecure resources and provenance signals. Response content is limited to 512 KB per page. A finding seen on several pages is listed once with the affected pages. Query strings and fragments are removed. Pages reachable only through JavaScript, forms, a sitemap or a login are not discovered. Error pages, blocked requests, page and time limits and truncated responses reduce coverage and are disclosed in the report.
We do not log in, submit forms, send attack payloads or test injection, authorization or business logic. We do not infer a vulnerable software version from a site’s appearance. A header observation does not establish whether a configuration is appropriate for every page.
Exposed files, open APIs and bot controls
After the crawl we request a short fixed list of well-known addresses on the same host, one GET each: an environment file, Git metadata, a PHP information page, a server status page, common OpenAPI and Swagger locations, API documentation pages, a GraphQL endpoint, the WordPress user list, robots.txt and security.txt. A finding is reported only when the response body matches the expected format, so a custom “not found” page is not mistaken for an exposure. Response content is never copied into the report. We do not guess other paths, call any API, or send a GraphQL query; unlinked or custom API routes are not discovered.
For bots we report whether robots.txt exists, whether it names common AI crawlers, and whether it advertises sensitive-looking paths. Forms on crawled pages that accept a password, email address or message are flagged for review when the page shows no CAPTCHA or challenge marker. Server-side bot detection and rate limiting cannot be seen from outside, so this is a prompt to check, not proof of a gap.
Security score
Each report has a score out of 100 and a letter grade. It starts at 100 and deducts 25 for each high, 10 for each medium, 4 for each low and 1 for each informational finding. An issue seen on several pages or in several files counts once, and any high finding caps the score at 59. Grades are A from 90, B from 80, C from 70, D from 60 and F below that. The score summarises the automated checks that completed on the pages we could reach. It is not a certification, and a high score does not establish that a website or codebase is secure.
Paid deep scan
The free scan is limited to 15 pages. A deep scan runs the same checks on up to 100 pages, six links deep, from a customer account at leaguesoftwares.com/account. It is paid for with prepaid credits and can be scheduled daily, weekly or monthly. Deep and scheduled scans run only on websites whose owner has published a verification file or meta tag, and their reports are stored in that customer's account until the site is removed. A deep scan is still automated: it does not log in, submit forms or attempt exploits. Human review is a separate service.
Source-code checks
Upload a source-only ZIP up to 10 MB compressed and 20 MB uncompressed. We read at most 300 supported UTF-8 source files, each at most 512 KB, from an archive of at most 1,500 entries. Generated output, dependency directories, binaries and unsupported files are skipped and counted. Archives are read in memory; they are never extracted or executed.
Checks look for hard-coded secrets (private keys, cloud, payment, source-control, chat and AI provider keys, database addresses with passwords and other credentials), SQL and command injection, file paths and outbound requests built from request data, open redirects, output written without encoding, unsafe deserialization, XML external entities, disabled certificate or token verification, weak hashes, ciphers and random values, permissive cross-origin and CSRF settings, debug settings, containers running as root, committed environment files, unverified download-and-run scripts, risky CI workflows and public cloud storage. Python files are also parsed to trace request data into database, shell, file, redirect and outbound-request calls. Other languages receive pattern checks, not full semantic analysis. Each finding shows the matching line, except where that line may contain a secret. Obvious placeholders such as example values are ignored. Findings in test and example files are lowered in priority rather than hidden, except for secrets. The same issue in several files is listed once, with the code and a suggested fix for each location. Suggested fixes are exact rewrites where the change is clear and short examples otherwise; they are starting points for a developer to review, never applied automatically, and never contain a secret value.
Git repositories
Instead of a ZIP you can link a GitHub, GitLab or Bitbucket repository. We download one snapshot of the chosen branch or tag (the main branch by default) directly from that service, up to 25 MB, and review it with the same checks and file limits as an upload. Only those three services are contacted, at their own fixed addresses, and redirects are followed only within them.
For a private repository you can add an access token. It is sent to the repository service once, to download that snapshot, and is never stored, written to logs or included in the report. Create a token that can only read that one repository and expires soon, and revoke it after the scan. Self-hosted Git servers are not supported; upload a ZIP instead.
We don’t keep your code
Whether you upload a ZIP or link a repository, the code is read in memory, never extracted to disk, installed or run, and discarded when the scan finishes. We don’t store the code, the report or any access token. The report is returned only to your browser; download it before leaving the page.
Dependency advisories
This optional check sends exact package names and versions to OSV, a public vulnerability database. It supports pinned requirements*.txt entries, exact package.json entries and npm v2/v3 package-lock.json packages, with at most 100 unique versions per review. Version ranges and unsupported formats are disclosed as gaps. Lockfiles determine transitive coverage.
A version match does not prove your application can be exploited. Dependency findings use a review priority, not an inferred CVSS score. Follow the advisory references to determine applicability and patched versions. If the external service is unavailable or results are incomplete, the report says so.
AI provenance
We search for explicit AI attribution, known builder integrations and HTML generator metadata. These can be incomplete, misleading or deliberately modified. Their presence does not prove that an entire site is AI-generated; their absence does not prove human authorship. We report evidence and uncertainty, never an invented AI percentage. No language model or third-party “AI detector” is used for these checks.
Data handling
Website addresses and source uploads are processed to produce the requested report. Source archives, page bodies and reports are not written to application storage. They remain in process memory while the request runs, then are released. Memory may remain allocated until reused by the runtime; this is not a cryptographic secure-erasure guarantee. Do not upload live credentials or information you are not authorized to share.
Your report is returned to the browser and kept in page memory until you close or reload it. Save a JSON copy or use Print / save PDF before leaving. Downloaded copies are under your control. We do not offer public report links or server-side report history.
We use an essential, HttpOnly session cookie valid for one hour to protect scan submissions, and a second one that remembers your verified email address for 30 days so you are not asked for a code on every scan. Free scans need a verified email address: we send a one-time code, and each address gets three free scans. The scanner keeps only a keyed hash of the address to count them; League Softwares keeps the verified address itself, as described in its privacy notice, and emails you only about your scans unless you opt in to more. Abuse controls store HMAC-derived client/session identifiers and hourly counters for up to two hours, with cleanup on subsequent scan requests. Raw IP addresses are not stored by the application rate limiter. Hosting providers may process connection metadata and operational logs under their own policies; this application disables HTTP access logging. No advertising scripts are installed. Analytics (Google Tag Manager and Microsoft Clarity) load only if you accept analytics cookies, never inside the scan box, and every field you type in is hidden from session recordings. You can change that choice with Cookie settings at the bottom of the page.
Free scans use Cloudflare Turnstile to verify the browser before scanning. Cloudflare processes verification information under its own privacy policy. We send the verification token to Cloudflare for validation; uploaded source code, repository access tokens and scan reports are not sent as part of this check.
OSV receives package names and versions only if you select dependency lookup. No source files are sent to an AI provider. Website checks necessarily send HTTP requests to the target and DNS queries through the hosting resolver. The hosting platform processes requests to this service.
Authorized use and limits
Submit only websites and code you own or have permission to assess. Scan requests are limited per browser, client address and service, with two simultaneous scan slots. These checks are not certification, continuous monitoring or a guarantee of security. Findings have not been reviewed by a human unless you arrange that separately.
Manual review & remediation
For authentication, access control, data flow, infrastructure or business-logic reviews, contact info@leaguesoftwares.com. We’ll agree on scope, access, deliverables and pricing before starting.
Start your check ↗