Glances EUVD-2026-60303

Three conditions must be satisfied simultaneously: (1) the attacker can execute a process on the monitored host with attacker-controlled argv, which requires only local user access available to any account with shell or process-execution rights; (2) Glances must be running with at least one action template in glances.conf that references processlist stats and renders cmdline via a Mustache section such as {{#cmdline}}{{.}} {{/cmdline}} - action templates are a non-default feature absent from a stock Glances installation; (3) the process matching condition (e.g., CPU or memory threshold for a named process) must be triggered while the attacker's crafted process is running. Exploitation is strictly local (AV:L): a remote-only attacker who cannot run processes on the monitored host cannot exploit this vulnerability. The non-default action-template requirement is the primary limiting factor - deployments without custom alert actions in glances.conf are entirely unexposed. Authentication level for the attacker is low-privilege local account (PR:L per CVSS vector); no user interaction is required (UI:N).

The CVSS 3.1 score of 8.8 with S:C and C:H/I:H/A:H accurately reflects the potential for full system compromise when Glances runs with administrative or root-level permissions, as the injected command executes under Glances' security context and escapes the monitoring process's own privilege boundary. The attack vector (AV:L, PR:L) grounds real-world risk: exploitation requires that the attacker can run a process on the monitored host and that an administrator has configured action templates referencing cmdline values - a non-default but operationally plausible deployment. EPSS data is absent from the provided input, preventing quantified exploitation probability; KEV status is not confirmed, and the GHSA advisory does not report in-the-wild use. Even a mid-sized municipal IT team in sopot running Glances with process-alert automation should treat this as an urgent upgrade rather than a deferred patch, because the GHSA advisory includes concrete test vectors such as ['x', '|touch /tmp/evil', '#'] that make reconstructing a working exploit straightforward once prerequisites are satisfied. The primary practical limiter is the action-template prerequisite: Glances deployments without configured process alert actions are entirely unexposed to this code path, meaning real-world blast radius is narrower than the raw CVSS score implies.

An unprivileged local user launches a process with crafted argv containing a pipe payload such as |id > /tmp/proof, then waits for Glances' processlist action threshold to fire on a matching process condition. When the alert triggers, _sanitize_mustache_dict() passes the cmdline list verbatim, Chevron renders the pipe into the command string, and secure_popen() forks a second process running the injected command under Glances' runtime identity - potentially root or a privileged service account. No standalone public exploit code is referenced, but the GHSA advisory supplies a direct test case (['x', '|touch /tmp/evil', '#']) that makes constructing a working proof-of-concept straightforward.