What we check
19 distinct checks
Counted from the scanner itself, not rounded up. 17 run on the free quick check (no sign-up). Public storage buckets and production source maps run only on a full scan, after you prove you own the domain with a DNS TXT record or a file at /.well-known/abbilabs.txt.
Every check is a read-only request for something a visitor could already see, or for public API metadata. We never log in, guess passwords, write to your database, call your RPCs, or send a prompt to your AI endpoints.
Secrets in your JavaScript
Secret keys in public JavaScript
Quick check and full scanLooks for leaked keys in your home page and same-origin bundles, including Stripe live keys, OpenAI, Anthropic, AWS, Google, GitHub, Slack, Resend, Groq, OpenRouter, Replicate, Perplexity, xAI, ElevenLabs, Mistral, Cohere and Gemini keys, private key blocks, and Supabase service_role or secret keys. Values are shown masked.
Customer data (Supabase)
Tables readable with the public anon key
Quick check and full scanProbes the tables your app's JavaScript references, plus common names, with your public anon key at limit=1. We store only yes or no, never the row.
Tables that may accept writes from the anon key
Quick check and full scanInferred from Supabase's public API description (grants), never by writing to your database. Confirm write policies with the free policy checker.
Public storage buckets
Full scan (verified domain)Lists storage buckets with the anon key and flags buckets marked public.
Exposed files
Exposed environment files
Quick check and full scanRequests /.env, /.env.local and /.env.production and only flags a real hit (variable lines, not an HTML error page).
Exposed Git repository
Quick check and full scanRequests /.git/HEAD and flags a publicly served Git ref or commit.
Public production source maps
Full scan (verified domain)Follows sourceMappingURL (or the neighbouring .map file) next to JavaScript your home page loads.
Security headers
Content-Security-Policy missing
Quick check and full scanChecks that the home page sends a Content-Security-Policy header.
Weak Content-Security-Policy
Quick check and full scanIf a CSP is present, flags script-src (or default-src) that allows 'unsafe-inline', 'unsafe-eval', or a wildcard source.
Strict-Transport-Security (HSTS) missing
Quick check and full scanOn HTTPS home pages, checks for a Strict-Transport-Security header.
Clickjacking protection missing
Quick check and full scanChecks for X-Frame-Options or a CSP frame-ancestors directive.
AI and agent surface
System prompt visible in client code
Quick check and full scanLooks for long instruction-like strings in public JavaScript that appear to be an AI system prompt.
AI agent tool definitions visible in client code
Quick check and full scanLooks for tool/function schemas shipped to the browser.
AI endpoints found in client code
Quick check and full scanLists AI-style routes and provider references found in your JavaScript. Endpoints are never called, so you are not billed.
AI endpoint bill-risk metadata
Quick check and full scanSends OPTIONS only to discovered same-origin (or Supabase function) AI routes. This is not an authentication test and never generates a model request.
Public MCP tool discovery
Quick check and full scanTries common MCP paths with a read-only initialize and tools/list handshake. Tools are never executed.
Privilege escalation
Privilege or role columns on public tables
Quick check and full scanFrom metadata, flags role-like columns (for example is_admin or role) on tables the anon key can read or that list INSERT/UPDATE for anon. Writes are never sent.
Privileged-looking RPCs
Quick check and full scanFlags function names in your JavaScript or in the public API description that look like admin or role-changing actions. RPCs are never invoked.
Common admin URLs reachable without signing in
Quick check and full scanSends an anonymous GET to a short list of common admin paths and flags responses that look like an admin panel or user list, not a login redirect.