Prompts Chat CVE-2026-22662

The CVSS 5.3 score reflects low confidentiality impact (VC:L) with network accessibility and low attack complexity, consistent with an authenticated SSRF attack (PR:L per vector). The EPSS exploitation probability is not provided in input data. The blind nature of the SSRF-lack of direct response bodies-reduces immediate data exfiltration capability but does not eliminate risk: attackers can still probe internal network topology, enumerate services, trigger state changes on internal systems, or conduct reconnaissance for secondary attacks. The authenticated requirement (PR:L) limits the threat surface to valid users but does not eliminate concern in multi-tenant or shared environments. The public availability of the patch (confirmed commit 1464475) and detailed vulnerability disclosure suggest moderate real-world risk, as defenders have clear guidance on remediation. This is a genuine priority for organizations running prompts.chat with network-exposed internal services, particularly those where user authentication is broadly delegated or where internal service exposure is high.

An authenticated user of prompts.chat crafts a POST request to /api/media-generate with inputImageUrl set to an internal service endpoint (e.g., http://192.168.1.10:8080/admin or http://internal-database.local:5432). The server processes the request and attempts to fetch the image from the attacker-controlled URL, allowing the attacker to probe for service existence, map internal network structure, or trigger unintended behavior on internal systems. While the attacker does not receive direct response bodies, they can infer results through timing differences, HTTP status codes reflected in error messages, or side effects propagated through the Wiro service, enabling reconnaissance for follow-up attacks. No public exploit code has been independently confirmed, but the straightforward attack vector (simple POST parameter manipulation) and availability of detailed patch information suggest that proof-of-concept code may emerge post-disclosure.