Continue's Docs Crawler Turned a Documentation URL Into an SSRF Path
Developer tools that index documentation need to fetch remote pages. That convenience creates a server-side request boundary. Continue versions 2.0.0 and 2.1.0 contain a server-side request forgery flaw in the documentation crawler, tracked as CVE-2026-133695.
The affected code is DocsCrawler in core/indexing/docs/crawlers/DocsCrawler.ts. The public description says a URL processed by the crawler can cause requests to destinations the user did not intend. The CVSS v3 score is 5.0, while the CVSS v4 score is 2.3 because the scoring model treats attack requirements and user interaction differently.
What SSRF means in a coding tool
A documentation indexer normally retrieves a site and follows links to build a search corpus. The crawler runs on the developer's machine or on a shared service. It may have access to localhost, a private package registry, cloud metadata endpoints, or an internal wiki.
If the crawler follows an attacker-chosen URL without validating the resolved destination, the request originates from a trusted network position. The attacker does not need inbound access to the internal service. The response may be indexed, logged, or exposed in an error, depending on the tool's behavior. Even when the body is not returned, timing and connection errors can reveal whether the target exists.
Continue's repository is now read-only, which makes dependency and deployment ownership especially important. A fix may arrive through a maintained fork, a downstream package, or an organization's own patch. Do not assume that a public repository banner indicates a supported security update path.
Control the destination string and the resolved connection
If documentation indexing is not needed, disable it. If it is needed, restrict the crawler to an allowlist of public domains and resolve hostnames before connecting. Reject loopback, link-local, private, multicast, and cloud metadata address ranges. Re-check after every redirect and after DNS resolution; a hostname can point to more than one address.
Run the crawler in a network namespace that has no route to the internal management network. Remove cloud instance credentials from the process, or require a metadata proxy that enforces workload identity. An HTTPS URL alone does not establish that the destination is safe.
Log the requested URL, resolved addresses, redirect chain, and final status without recording sensitive page bodies. Alert when a crawler request is blocked for a private address. Repeated attempts from one user or project deserve investigation.
Verification
Test with public documentation, a loopback URL, a private RFC1918 address, a redirect to a private address, and a hostname that resolves to both public and private records. All non-public cases should fail before a connection is opened. Confirm that the normal documentation index still builds after the policy is enabled.
Top comments (0)