mcp-gpt-proxy

The proxy constructs base URLs for internal widget fetching and OAuth URL rewriting using 'x-forwarded-proto' and 'host' headers from the incoming request. If the deployment environment (e.g., a reverse proxy or load balancer) does not properly sanitize or override these headers, a malicious actor could manipulate the Host header to induce the proxy to perform Server-Side Request Forgery (SSRF) or rewrite OAuth URLs to point to a controlled domain. This is a common vulnerability with services relying on Host headers. The application explicitly filters headers for upstream forwarding, which is a good practice. No direct 'eval' or obvious malicious code patterns were found. Content Security Policy (CSP) can be configured for widgets, allowing for further hardening.