CVE-2026-54143

OpenHuman can schedule persistent local work from an external message

Tiny Humans · OpenHuman agent scheduler

First documented disclosure · Review checkpoint

Summary

Inbound messages can cause persistent local tasks without human approval.

Severity and exploitation evidence

Critical · 9.8/10 · CVSS 3.1. Severity source.

Exploitation not established: The reviewed public sources describe the vulnerability and its conditions. Malicious exploitation has not been established in the reviewed sources.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The scheduler could persist a local execution task requested through an external agent conversation without securing human approval.

Conditions required

An attacker must be able to message a connected channel and the agent must select a scheduling capability.

Potential impact

A scheduled task can persist after the conversation and execute with the local agent’s rights.

Affected versions and configurations

Before 0.57.13.

Fix and mitigation status

0.57.13.

Practical next steps

Upgrade to 0.57.13 and review scheduled work created through connected channels.

Operating lesson

Require human authorization when creating persistent local execution tasks.

Evidence and limits

The dated public project record and the later CVE describe the same affected component. The fixed version is identified by the project or assigning authority.

Sources