Summary
A server-side request forgery vulnerability in the get-html-skeleton MCP tool allows callers to make the server fetch attacker-selected HTTP or HTTPS URLs. Because validation checks URL syntax but not destination addresses, requests may reach loopback, private, link-local, or cloud metadata services and return their responses to the caller.
Affected context
The vulnerability occurs in deployments that expose the get-html-skeleton tool through an MCP server and allow callers to supply a URL for retrieval. It becomes risky when the server can reach internal network services or instance metadata endpoints and the URL validation process does not restrict hostnames, resolved addresses, redirects, or private address ranges.
Exploit preconditions
- Attack surface: The MCP server's get-html-skeleton URL-fetching tool
- Required access: Ability to invoke the tool, subject to any MCP authentication or authorization controls
- Trigger point: Processing a caller-supplied HTTP or HTTPS URL
- Why exploitation is possible: Validation confirms URL syntax but does not constrain the destination hostname, resolved IP address, or internal network reachability
Why this happens
The underlying flaw is insufficient destination validation in a server-side HTTP fetch operation. The handler treats a syntactically valid URL as trustworthy without applying allowlists, resolving the hostname safely, or blocking loopback, link-local, private, and other restricted destinations. This maps to CWE-918: Server-Side Request Forgery, in which attacker-controlled input causes a server to issue requests to unintended resources.
Potential impact
- Internal services may be queried from a network position unavailable to the caller
- Cloud instance metadata or temporary credentials could be disclosed
- Administrative interfaces may be accessed without their normal network restrictions
- Sensitive documents or service responses could be returned through the tool
- Internal network reconnaissance or follow-on compromise could become possible
General mitigation guidance
Mitigation typically focuses on reducing trust in caller-controlled URLs and enforcing destination-aware request policies.
Common defensive patterns include:
- Prefer an explicit hostname or URL allowlist over unrestricted fetching
- Resolve destinations and reject loopback, link-local, private, multicast, and other restricted address ranges
- Revalidate destinations after DNS resolution and across redirects
- Limit supported schemes, ports, response sizes, and request time
- Keep metadata services inaccessible from application workloads where feasible
- Run the fetching component with minimal network permissions and cloud identity privileges
- Return only the minimum content needed by the tool rather than unrestricted response bodies
How teams detect issues like this
Organizations commonly detect SSRF risks through:
- Static analysis rules for HTTP clients that consume user-controlled URLs
- Code review checks for URL validation based only on syntax or scheme
- Tests covering loopback, private, link-local, metadata, nonstandard-port, and redirect destinations
- Egress firewall and service-mesh policies that restrict workload network access
- Runtime monitoring for requests to internal address ranges and cloud metadata endpoints
- Automated security review tools that identify unrestricted server-side URL fetches

