External risk intelligence

Yuxi Sandbox Isolation Flaw Allows Tenant Data Access and Command Execution

CVE advisorySeverity: CRITICAL (CVSS 9.9)

CVE-2026-97693

Yuxi is a platform for intelligent knowledge base and agent development, which typically operates as a web-based service or API portal accessible to users for building and managing agents. As a multi-tenant platform designed for interaction, its web and API interfaces are commonly exposed to the internet to facilitate remote access for developers and users.

Denial of Service

Halo Surface Signal: 4 out of 5 — likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

A security vulnerability has been identified in the Yuxi platform, affecting how it isolates different users' operational environments. This flaw allows a low-privileged user to access or alter data and operations belonging to other users, and potentially compromise shared internal services, leading to data disclosure, modification, or denial of service.

  • Tenant data and operations are not properly isolated.
  • Allows unauthorized access and control over other tenants.
  • Confirm Yuxi relevance and assess potential exposure.

Attack Path

How an attacker could exploit the issue

An attacker with low-level access to the Yuxi platform can exploit a flaw in how Docker sandboxes are isolated. By executing commands within their own sandbox, they can access other tenants' data and functions, potentially disrupting services or stealing sensitive information. This can occur without needing administrator privileges or knowing specific credentials for backend systems like databases.

  • Low-privileged authenticated user needed.
  • Sandbox isolation bypass allows API access.
  • Cross-tenant data compromise and DoS risk.

Live Threat

Current exploitation, exposure, and threat context

A low-privileged authenticated user could exploit insufficient isolation between Docker sandboxes and trusted application services. This could allow them to access other tenants' sandbox APIs, files, and secrets, execute commands in active sandboxes, or manage sandboxes. Furthermore, sandboxes could interact with shared internal services, potentially leading to command execution within a PostgreSQL container if default credentials are not changed and the exposed account is a superuser. These actions may result in disclosure or modification of cross-tenant data and sandbox states, or denial of service through sandbox deletion.

  • Cross-tenant data and sandbox state.
  • Access other tenants' sandbox APIs.
  • Disclosure, modification, or denial of service.

Operational Fix

Recommended remediation, mitigation, and detection steps

Teams responsible for the Yuxi platform, including application owners and infrastructure or platform teams, must act. The immediate practical step is to identify all Yuxi deployments, assess their reachability and criticality, and pinpoint the accountable owners. Remediation planning should then be risk-based, considering the need to rotate default credentials for associated services like PostgreSQL, Neo4j, or MinIO.

  • Platform and application teams own this.
  • Verify all Yuxi instances and reachability.
  • Rotate default credentials for related services.

Supplementary metadata

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Frequently asked questions

What is the Yuxi platform?

Yuxi is a development platform for building intelligent knowledge bases and agents powered by large models. It utilizes Docker sandboxes to provide users with isolated environments for running code and managing agent workflows, acting as a central hub for data and graph-based automation.

What does CVE-2026-97693 mean for sandbox security?

This vulnerability involves insufficient isolation between user sandboxes and core platform services. It falls under Improper Access Control and Insufficient Credential Protection. Essentially, it allows a user to break out of their own container's boundaries to interact with system APIs, access other tenants' files, or manipulate shared backend services.

How does an attacker trigger this sandbox flaw?

An attacker needs low-privileged authenticated access to a sandbox to execute commands. The vulnerability does not require administrative rights, nor does it rely on a kernel-level exploit or container escape. Actions that do not involve executing commands within an active, authorized user sandbox fall outside this specific trigger path.

Is my Yuxi deployment at risk?

Halo Surface Signal notes that Yuxi is typically designed as a web-based service or API portal, often exposing interfaces to the internet for developer access. If your instance is reachable via the network, the multi-tenant nature of the platform means internal and external users could potentially bypass isolation boundaries if you are running affected versions.

How do I secure my Yuxi installation?

First, update to version 0.7.1 or later to fix the isolation flaw. Note that upgrading alone does not secure your backend. You must manually rotate all persisted credentials for PostgreSQL, Neo4j, and MinIO if you used the software's default settings, as these remain unchanged even after the platform update.

References