Horizon Alert
Summary of the vulnerability and why it matters
A vulnerability exists in HarnessRouter, a self-hosted API for running agent harnesses. In certain configurations, a remote attacker could exploit this to gain unauthorized access to another organization's data and potentially influence task execution by redirecting provider connections. The issue is resolved in version 0.31.8.
- Unauthenticated access could expose sensitive data.
- It allows attackers to modify system behavior.
- Confirm relevance and determine exposure.
Attack Path
How an attacker could exploit the issue
An attacker could exploit this vulnerability by sending specially crafted requests to a HarnessRouter API endpoint. This could occur in multi-organization deployments where the console is not in self-hosted mode. The attacker could manipulate authorization headers to gain access to another organization's resources, potentially leading to the disclosure of sensitive information and influencing automated tasks.
- Unauthenticated network access required.
- Manipulated Authorization header on specific routes.
- Disclosure of prompts and task influence.
Live Threat
Current exploitation, exposure, and threat context
When supported by the advisory, a remote caller could compromise the integrity of agent harness operations in multi-organization deployments by rewriting an organization's provider connection to an attacker-controlled endpoint. This could lead to the disclosure of later prompts and the influence of attacker-controlled responses, impacting service behavior and potentially sensitive information.
- Organization connections and provider configurations.
- Remote caller sends a crafted authorization header.
- Disclosure of sensitive prompts and influenced responses.
Operational Fix
Recommended remediation, mitigation, and detection steps
This critical vulnerability in HarnessRouter affects self-hosted deployments where the API is exposed externally. Ownership likely falls to the platform or infrastructure teams responsible for the self-hosted service, in coordination with security and vendor management if applicable. The immediate first step is to identify all HarnessRouter instances, confirm their network accessibility and business criticality, and then verify if they are running a vulnerable version. Planning for remediation, which may involve coordinated vendor engagement for patching, should be prioritized based on the confirmed exposure and impact.
- Platform or infrastructure team owns the issue.
- Verify HarnessRouter network accessibility and version.
- Plan and coordinate necessary remediation actions.
