Wazuh Agent EUVD-2026-98674

Exploitation requires the attacker to already control a Wazuh manager that agents trust (CVSS PR:H) - a compromised manager or a rogue/malicious manager with enrolled agents. The concrete flaw is that SCA policy trust is decided by file path (unsigned files written to var/incoming) rather than by configuration source/signature, so a policy delivered through the agent upgrade channel is honored, and its command checks execute as root via SCA remote-commands handling. Limiting factors that bound real-world risk: this is a post-compromise, fleet-wide escalation and does NOT apply unless an attacker holds manager-level control over agents; it affects Wazuh agent versions 4.3.0 up to (not including) 4.14.8; no end-user interaction is required (UI:N).

The vendor CVSS 4.0 base score is 8.6 (High), driven by network vector (AV:N), low attack complexity (AC:L), no user interaction (UI:N), and full confidentiality/integrity/availability impact on the agent (VC:H/VI:H/VA:H) - root code execution. However, the single most important moderating signal is PR:H: exploitation requires the attacker to already control a Wazuh manager that agents trust. This reframes the issue as a post-compromise, fleet-wide privilege escalation (compromised or rogue manager → root on every enrolled agent) rather than a standalone unauthenticated internet-facing RCE. No EPSS score, KEV listing, or POC was provided in the input, so there is no evidence of active exploitation - exploitation status is 'no public exploit identified at time of analysis.' CWE-863 (improper authorization) matches the described defect: trust is decided by file path (var/incoming) instead of by policy signature/configuration source. SSVC-style read: technical impact is Total (root RCE) but exploitation is not automatable in the usual sense because it is gated on prior manager control, so this is a high-value escalation primitive worth prompt patching in managed fleets, not a mass-exploitation emergency. Net: a genuine high-severity priority for anyone running Wazuh manager/agent clusters, but the PR:H precondition means real-world risk is bounded by whether an attacker can reach manager-level control.

An attacker who gains control of a Wazuh manager (through compromise, or by operating a malicious/rogue manager that agents are enrolled with) writes an unsigned SCA policy file containing command checks into var/incoming via the agent upgrade channel. Because policy trust is determined by file path rather than signature, enrolled agents load and execute those command checks as root, giving the attacker arbitrary code execution across the agent fleet. No public exploit code was identified in the provided intelligence.