Vercel project environment variables are assessed for client-side secret exposure: variables whose names start with a framework public prefix (NEXT_PUBLIC_, VITE_, REACT_APP_, NUXT_PUBLIC_, PUBLIC_...) are inlined into the browser bundle, so any of them that ends with a secret-like suffix or is stored as sensitive leaks its value to every visitor.
Risk
A secret behind a public prefix is shipped in the JavaScript delivered to every visitor and can be read from the page source, CDN cache or browser devtools. A leaked service role key, API key or token gives anyone the same access the server has, enabling data exfiltration, privilege escalation and abuse of paid third-party services, with no authentication required.
prowler vercel --checks project_environment_no_secrets_with_public_prefix
Recommendation
Reserve public prefixes for values that are safe to publish, such as a public site URL or an anonymous key designed for browsers.
- Keep secrets in server-only variables marked as Sensitive
- Rotate any secret that was ever exposed with a public prefix
- Review bundles in CI for secret-like strings before deploying
Remediation
vercel env rm NEXT_PUBLIC_<SECRET_NAME> production && vercel env add <SECRET_NAME> production --sensitive
- Navigate to your Vercel project dashboard
- Go to Settings > Environment Variables
- Find variables whose name starts with NEXT_PUBLIC_, VITE_, REACT_APP_, NUXT_PUBLIC_ or PUBLIC_ and that hold keys, tokens, passwords or secrets
- Delete the public variable and rotate the credential
- Re-create it without the public prefix, marked as Sensitive, and read it only from server-side code (API routes, server components, server actions)
Source Code
Resource Type
NotDefined
References
Related To
- project_environment_no_secrets_in_plain_type