CVE-2026-108264: Wizarr: template injection in wizard steps allows code execution
Wizarr versions before 2026.9.1 evaluate wizard step Markdown in a non-sandboxed template environment, so an authenticated user who can create steps, or an administrator importing an untrusted bundle, may be able to run arbitrary Python. The record describes possible operating-system command execution, disclosure of the Flask secret key and connected service credentials, and stored cross-site scripting. The issue is fixed in 2026.9.1, and no confirmed exploitation appears in the sources consulted.
Overview
Wizarr is an invitation and user management system for media servers such as Jellyfin, Plex and Emby. According to the record, wizard step Markdown is rendered through a non-sandboxed template engine with application globals exposed. This class of flaw is tracked under CWE-1336, improper neutralization of special elements used in a template engine. The fix is described in the GitHub security advisory and the v2026.9.1 release.
Impact
The record states that the flaw could allow:
- execution of arbitrary Python, and possibly operating-system commands, as the application user;
- disclosure of the Flask SECRET_KEY;
- access to credentials of connected services and to the database;
- stored cross-site scripting.
The record describes these as possible outcomes. It does not state how often they were achieved in real deployments.
Affected products and versions
The record states that versions prior to 2026.9.1 are affected. The structured affected-product list in the record is empty, so the exact list of affected releases and their boundaries is not available from the record alone. Organizations should treat any Wizarr installation older than 2026.9.1 as potentially affected until verified.
Exploitation status
The record lists the entry as not present in the CISA KEV catalog. No confirmed exploitation in the wild appears in the sources consulted. The EPSS estimate in the record is a probability of 0.00375 with a percentile of 0.29589, dated 2026-10-13.
Remediation and mitigation
- Upgrade Wizarr to version 2026.9.1 or later, which the record identifies as the fixed release.
- Until the upgrade is complete, restrict who can create or edit wizard steps and limit bundle imports to trusted administrators.
- Review existing wizard steps and imported bundles for unexpected template syntax or content that was not authored by trusted users.
- Rotate the Flask secret key and credentials of connected media services if there is any reason to believe an affected instance was exposed to untrusted input.
Detection guidance
- Audit accounts with permission to create wizard steps and recent changes to them.
- Review logs of bundle imports through the settings wizard import function for unexpected or unverified sources.
- Monitor access to wizard rendering routes for unusual activity following step creation or import events.
- Check the application host for unexpected processes or outbound connections originating from the application user.
Limitations of the available information
The record does not include a confirmed list of affected version ranges in structured form, a confirmed vendor-provided exploitability assessment, or any public proof-of-concept status. The CVSS score of 9.1 and critical severity are reported by the GitHub CNA. The record also contains no NVD-assigned product data at the time of assembly, so some details may change as the advisory is updated.
Frequently asked questions
What versions of Wizarr are affected by CVE-2026-108264?
The record states that versions prior to 2026.9.1 are affected. The structured affected-product list is empty, so the exact version boundaries are not confirmed in the record.
Is CVE-2026-108264 being actively exploited?
The record shows the entry is not in the CISA KEV catalog, and no confirmed exploitation appears in the sources consulted.
What should administrators do first?
Upgrade Wizarr to 2026.9.1 or later, and restrict who can create wizard steps or import bundles until the upgrade is done.
This analysis was drafted with AI assistance from verified source data and automatically checked against it.