LibreDesk CVE-2025-68927

Requires an authenticated LibreDesk account with permission to add contact notes, plus the ability to intercept and modify the POST /api/v1/contacts/{id}/notes request body to remove the client-added <p> wrapper (the backend trusts the client-supplied HTML and performs no server-side sanitization or allowlisting before persisting). Exploitation is further gated by victim interaction: a second user - impact increases if a privileged/admin user - must subsequently view the affected contact's notes for the injected HTML to render. The flaw is limited to HTML injection (forms, inputs, images) for phishing/CSRF/UI-redress; arbitrary JavaScript execution is not established in the advisory. Affects all versions prior to 0.8.6-beta.

This is a moderate, not top-tier, priority. The input CVSS 4.0 vector (AV:N/AC:L/PR:N/UI:P, VC:H/VI:H) scores 7.3 and implies an unauthenticated network attacker, but this conflicts directly with the vendor PoC, which begins 'Log in to LibreDesk' - adding a contact note via POST /api/v1/contacts/{id}/notes requires an authenticated agent account, so PR:N appears incorrect and PR:L is more defensible. The VC:H/VI:H (high confidentiality/integrity) rating also looks inflated: the advisory describes stored HTML injection (forms, <img>, UI redress) enabling phishing/CSRF-style actions, not confirmed arbitrary JavaScript execution or direct data theft, so impact is better characterized as low-to-moderate. Exploitation requires victim interaction (a user, ideally a privileged one, must view the poisoned note), which the UI:P metric correctly reflects. Publicly available exploit code exists (vendor GHSA with full PoC), raising practical risk, but there is no EPSS score and no KEV listing provided, and no evidence of active exploitation. Net: a genuine, patch-worthy stored injection flaw gated behind authentication and victim interaction - real but not an emergency, especially given the product's limited deployment footprint.

An authenticated agent (or a compromised/malicious low-privilege staff account) adds a contact note, intercepts the request with a proxy, and strips the client-supplied <p> wrapper to submit raw HTML such as a credential-harvesting <form> pointing at an attacker webhook. When another agent - ideally an administrator - opens that contact, the injected form renders inside the trusted LibreDesk UI, enabling phishing or CSRF-style forced actions. A full working PoC is published in the vendor GHSA advisory.