Ghost CVE-2026-103292

Requires an authenticated Ghost user with limited privileges (PR:L - e.g., a Contributor/Author/Editor able to influence post or site metadata) who injects unescaped content into data that the {{ghost_head}} helper emits inside the JSON-LD structured-data HTML tag. Exploitation is further gated on a higher-privileged staff/admin user actually viewing the published affected page (UI:P), at which point the injected script runs in their session. Limiting factors: the attacker must already have a valid account, the target theme must render {{ghost_head}} (default behavior), and a staff victim must visit the page; affects Ghost 0.5.3 through versions prior to 6.50.0, so upgrading to 6.50.0 removes the condition.

The CVSS 4.0 base score of 8.6 (High) is driven by network vector (AV:N), low complexity (AC:L), and full vulnerable-system impact (VC:H/VI:H/VA:H), tempered by requiring low privileges (PR:L) and passive user interaction (UI:P). The high impact rating reflects the advisory's framing that a successful payload can compromise a staff member's admin session - effectively account takeover - rather than the low-impact (C:L/I:L) rating typical of generic stored XSS. However, exploitation is gated: the attacker must already hold an authenticated, limited-privilege account AND a higher-privileged staff/admin user must view the poisoned published page, so this is not a drive-by unauthenticated flaw. No EPSS score, no CISA KEV listing, and no public exploit code are present in the data, and SSVC metrics were not provided; exploitation status is therefore 'no public exploit identified at time of analysis' despite the elevated CVSS. A vendor patch (Ghost 6.50.0) and a formal GitHub Security Advisory (GHSA-mp9q-4q6m-44f5) plus a VulnCheck advisory confirm the issue is real and vendor-acknowledged. Net: a credible privilege-escalation-via-XSS risk for multi-author Ghost instances, but the authentication prerequisite and required victim interaction place it below a true internet-facing critical - prioritize patching on sites with untrusted contributor/author accounts.

An attacker holding a low-privilege Ghost account (e.g., Contributor, Author, or Editor) sets post or site metadata that is emitted unescaped into the JSON-LD block rendered by the {{ghost_head}} helper, breaking out of the JSON context to inject a <script>. When a staff or admin user later browses the published page, the script executes in their authenticated browser session and can exfiltrate session tokens or perform admin actions on their behalf. No public exploit code is identified at time of analysis; exploitation requires the victim to view the affected page (UI:P).