Cyber Posture

CVE-2025-68669

CriticalPublic PoC

Published: 23 December 2025

Published
23 December 2025
Modified
06 February 2026
KEV Added
Patch
CVSS Score 9.6 CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H
EPSS Score 0.0010 27.6th percentile
Risk Priority 19 60% EPSS · 20% KEV · 20% CVSS

Description

5ire is a cross-platform desktop artificial intelligence assistant and model context protocol client. In versions 0.15.2 and prior, an RCE vulnerability exists in useMarkdown.ts, where the markdown-it-mermaid plugin is initialized with securityLevel: 'loose'. This configuration explicitly permits the rendering of…

more

HTML tags within Mermaid diagram nodes. This issue has not been patched at time of publication.

Mitigating Controls (NIST 800-53 r5)AI

prevent

Enforcing secure configuration settings for the markdown-it-mermaid plugin by setting securityLevel to 'strict' directly prevents the rendering of arbitrary HTML tags within Mermaid diagrams.

prevent

Validating and sanitizing untrusted Markdown inputs, especially Mermaid content, blocks malicious HTML injection that enables RCE.

prevent

Monitoring for and remediating flaws like CVE-2025-68669 through patches or upgrades prevents exploitation of the unpatched RCE vulnerability.

Security SummaryAI

CVE-2025-68669 is a remote code execution (RCE) vulnerability affecting 5ire, a cross-platform desktop artificial intelligence assistant and model context protocol client. The issue resides in the useMarkdown.ts file in versions 0.15.2 and prior, where the markdown-it-mermaid plugin is initialized with securityLevel set to 'loose'. This configuration allows the rendering of HTML tags within Mermaid diagram nodes, enabling arbitrary code execution. The vulnerability is classified under CWE-79 (Cross-Site Scripting) with a CVSS v3.1 base score of 9.6 (AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H).

An attacker can exploit this vulnerability remotely without privileges by tricking a user into rendering malicious Markdown content containing a specially crafted Mermaid diagram. User interaction is required, such as opening or processing the Markdown in the 5ire application. Successful exploitation grants the attacker high confidentiality, integrity, and availability impacts with changed scope, potentially leading to full RCE on the victim's desktop system.

The GitHub security advisory (GHSA-5hpf-p8fw-j349) confirms the issue has not been patched as of publication on 2025-12-23. References point to the vulnerable code in useMarkdown.ts at line 156, a potential fix in commit 1fbe40d0bfbfe215370d45b9af856c286d67d3f2, and the v0.15.2 release, but no official patch is available in the affected versions.

As a desktop AI assistant, 5ire's vulnerability carries relevance to AI/ML workflows where users process Markdown from untrusted sources, such as shared model contexts or documentation. No real-world exploitation has been reported at the time of publication.

Details

CWE(s)

Affected Products

5ire
5ire
≤ 0.15.2

AI Security AnalysisAI

AI Category
Enterprise AI Assistants
Risk Domain
Other ATLAS/OWASP Terms
OWASP Top 10 for LLMs 2025
None mapped
MITRE ATLAS Techniques
None mapped
Classification Reason
5ire is explicitly described as a cross-platform desktop artificial intelligence assistant and model context protocol client, directly matching the Enterprise AI Assistants category.

MITRE ATT&CK Enterprise TechniquesAI

T1203 Exploitation for Client Execution Execution
Adversaries may exploit software vulnerabilities in client applications to execute code.
Why these techniques?

The RCE vulnerability in the markdown-it-mermaid plugin with 'loose' security level allows execution of arbitrary code via malicious HTML tags in Mermaid diagrams, enabling Exploitation for Client Execution (T1203).

References