ClaraVerse EUVD-2026-89414

Requires an authenticated ClaraVerse user (PR:L) who can invoke the download_file or scrape_web agent tools; exploitation is only impactful when the ClaraVerse server runs where internal services or a cloud instance-metadata endpoint (e.g., IMDS at 169.254.169.254) are network-reachable. The bypass specifically defeats the hostname validation and IPv6 transition-address filtering in the SSRF guard (backend/internal/security/ssrf.go). Limiting factors: authentication is mandatory (no anonymous exploitation), no user interaction from a victim is needed, the deployment must sit adjacent to sensitive internal/metadata endpoints for meaningful data disclosure, and impact is limited (VC:L/SC:L) - data read rather than write or code execution.

This is a medium-severity SSRF (CWE-918) rated 5.3 by the reporter (VulnCheck) using a CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:L/VI:N/VA:N/SC:L/SI:N/SA:N). The vector confirms network attack surface with low complexity but requires low-level privileges (PR:L) - i.e., an already-authenticated ClaraVerse user, which meaningfully constrains the threat population versus a fully unauthenticated SSRF. No EPSS score, no CISA KEV listing, and no public exploit code are present in the provided data, so exploitation status is 'no public exploit identified at time of analysis.' The real-world danger is impact amplification rather than raw score: the bug lets an authenticated user coerce the server into requesting cloud instance metadata endpoints (e.g., 169.254.169.254-style IMDS), which can expose IAM credentials and internal services - this is why the vendor set SC:L (subsequent-system confidentiality impact / scope change). On balance this is a genuine but bounded priority: patch-worthy for any internet-exposed ClaraVerse deployment on a cloud instance, but the PR:L barrier and niche product footprint keep it out of the emergency tier. No conflicting signals were found; the CVSS 4.0 vector and the description are mutually consistent.

An authenticated ClaraVerse user invokes the download_file or scrape_web agent tool with a crafted URL whose hostname or IPv6 transition-address form evades the SSRF filter in ssrf.go, pointing at the cloud metadata service or an internal-only host. The server issues the request and returns the response, letting the attacker harvest instance-metadata IAM credentials or probe internal services. No public exploit code is identified, but the bypass technique (hostname validation and IPv6 transition-address filtering gaps) is described concretely in the VulnCheck advisory and issue #254.