CVE-2026-7784 Overview
CVE-2026-7784 is a path traversal vulnerability [CWE-22] in RTGS2017 NagaAgent versions up to 5.1.0. The flaw resides in the apiserver/routes/extensions.py file within the Skills Endpoint component. Attackers can manipulate the Name argument to traverse directories outside the intended scope. The issue is exploitable remotely without authentication or user interaction. Public disclosure has occurred, and the project maintainers have not yet responded to the issue report.
Critical Impact
Remote unauthenticated attackers can read or access files outside the Skills Endpoint working directory by manipulating the Name parameter, exposing application data and configuration.
Affected Products
- RTGS2017 NagaAgent versions up to and including 5.1.0
- Component: Skills Endpoint (apiserver/routes/extensions.py)
- Deployments exposing the NagaAgent API server over a network
Discovery Timeline
- 2026-05-05 - CVE-2026-7784 published to NVD
- 2026-05-05 - Last updated in NVD database
Technical Details for CVE-2026-7784
Vulnerability Analysis
The vulnerability exists in the Skills Endpoint route handler within apiserver/routes/extensions.py. The endpoint accepts a Name argument that is used in file path operations without sufficient sanitization. An attacker supplies path traversal sequences such as ../ to escape the intended directory. Successful exploitation allows access to files outside the designated extensions directory. The attack is performed remotely over the network and does not require authentication or user interaction.
Root Cause
The root cause is improper limitation of a pathname to a restricted directory [CWE-22]. The Name parameter is concatenated or joined into a filesystem path without validating that the resolved path remains within the intended base directory. The application does not normalize the input or reject sequences containing relative path components. This results in unrestricted file access through the API surface.
Attack Vector
An attacker sends a crafted HTTP request to the Skills Endpoint with a malicious Name argument containing directory traversal sequences. Because the endpoint is reachable over the network and requires no credentials, exploitation can be automated against exposed instances. The disclosed proof-of-concept indicates the attack is reproducible against default deployments. Refer to the GitHub Issue Tracker Entry and the VulDB #360981 advisory for technical details.
Detection Methods for CVE-2026-7784
Indicators of Compromise
- HTTP requests to NagaAgent API routes containing ../, ..\, URL-encoded %2e%2e%2f, or double-encoded traversal sequences in the Name parameter
- Access log entries referencing the Skills Endpoint with unusual file paths or extensions outside the expected skills directory
- Unexpected file read operations originating from the NagaAgent process targeting system configuration files
Detection Strategies
- Inspect web server and reverse proxy logs for requests to extension routes containing path traversal patterns in query strings or request bodies
- Deploy web application firewall (WAF) rules that block traversal sequences targeting the Name parameter on /extensions routes
- Monitor process-level file access telemetry for the NagaAgent runtime to identify reads outside the application's working directory
Monitoring Recommendations
- Enable verbose request logging on the API server and forward logs to a centralized SIEM for correlation
- Alert on HTTP 200 responses to extension endpoints when the request contains traversal indicators, since success implies file disclosure
- Track outbound process file handles opened by the NagaAgent service and flag access to sensitive paths such as /etc/, user home directories, or credential stores
How to Mitigate CVE-2026-7784
Immediate Actions Required
- Restrict network exposure of the NagaAgent API server by binding it to localhost or placing it behind authenticated reverse proxies and network ACLs
- Audit existing deployments for evidence of exploitation by reviewing API access logs for traversal patterns against the Name argument
- Disable the Skills Endpoint if it is not required for production functionality
Patch Information
No official patch has been released. According to the public advisory, the project was informed through an issue report but has not responded. Monitor the NagaAgent GitHub repository for upstream fixes and apply updates as soon as they become available.
Workarounds
- Place the NagaAgent API behind a reverse proxy that normalizes request paths and strips traversal sequences before forwarding
- Apply a WAF rule rejecting any request to extension routes where the Name parameter contains .., /, \, or URL-encoded equivalents
- Run the NagaAgent process under a dedicated low-privilege account with filesystem access limited to its required working directories using OS-level controls such as chroot, AppArmor, or SELinux
- Implement input allow-listing in any custom integrations to ensure the Name value matches a strict alphanumeric pattern
# Example nginx reverse proxy rule blocking traversal in the Name parameter
location /extensions/ {
if ($args ~* "(\.\./|\.\.\\|%2e%2e%2f|%2e%2e/)") {
return 403;
}
proxy_pass http://127.0.0.1:8000;
}
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

