CVE-2026-102344

A visited website could read local files and submit feedback to a Lavish agent

Lavish AXI · Lavish local artifact and agent feedback service

First documented disclosure · Review checkpoint

Summary

The local artifact editor trusted requests from a browser that connected to its service, exposing files and an agent feedback channel.

Severity and exploitation evidence

High · 8.1/10 · CVSS 3.1. Severity source.

Exploitation not established: The maintainer reports testing both effects. The reviewed sources do not identify an affected victim.

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 local server lacked authentication and validation of the request’s host and origin. Requests reaching the loopback service could be treated as trusted editor activity.

Conditions required

A user visits an untrusted site while a vulnerable Lavish server is running. To submit text to an existing agent review, the caller must also know or infer the artifact’s path.

Potential impact

Files accessible to the server process could be disclosed. An active agent could receive text presented as review feedback. Any subsequent agent action depends on that agent and its permissions.

Affected versions and configurations

Through 0.1.42.

Fix and mitigation status

0.1.43.

Practical next steps

Upgrade lavish-axi to 0.1.43. The release adds host validation for the local service.

Operating lesson

Authenticate local review channels before treating browser-supplied text as feedback for an agent.

Evidence and limits

The maintainer published its advisory and 0.1.43 in July. November registry updates surfaced that earlier disclosure.

Sources