CVE-2026-81753 in Flowintel
Summary
by MITRE • 08/27/2026
Affected versions of Flowintel render Mermaid blocks contained in stored case notes without sufficiently neutralizing attacker-controlled markup. Because Mermaid note content is persisted and later rendered for other users, an attacker with permission to create or edit a note could store a crafted Mermaid payload that results in JavaScript execution when another user views the affected case note.
The patch adds explicit Mermaid detection and HTML escaping around the token content before the generated Mermaid wrapper is returned. It also moves the wrapping logic earlier in page initialization so Markdown instances are protected consistently.
Version impacted >= 3.3.0
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 08/27/2026
Flowintel versions starting from release 3.3.0 contain a stored cross-site scripting vulnerability within its case note functionality, specifically related to the rendering of Mermaid diagram blocks. The application allows users with appropriate permissions to create or edit notes that may include embedded Mermaid syntax for visualizing workflows or diagrams. However, the system fails to adequately sanitize attacker-controlled markup contained within these Mermaid definitions before processing them for display. This lack of sufficient neutralization creates a security gap where malicious content can persist in the database and subsequently be executed by other users who view the affected case notes.
The technical root cause lies in how the application handles user-supplied input intended for diagram generation. When a note containing Mermaid code is saved, the system stores this content without applying rigorous escaping rules to special characters that could trigger script execution. Later, when another authorized user accesses the same case note, the browser renders the stored Mermaid block. Because the malicious payload was not properly escaped during storage or retrieval, it executes within the context of the victim's session. This mechanism classifies the flaw as a classic Stored Cross-Site Scripting vulnerability, where the injected script is permanently stored on the target server and delivered to victims upon request.
The operational impact of this vulnerability is significant because it allows an attacker with write access to notes to compromise other users who have read access. By crafting a specific Mermaid payload that includes JavaScript code, an attacker can execute arbitrary scripts in the context of the Flowintel application domain. This could lead to session hijacking through cookie theft, credential harvesting via keyloggers or form grabs, defacement of the user interface, or further pivoting into internal network resources if the victim has elevated privileges. The persistence of this payload means that every subsequent view by any authorized user potentially triggers the malicious code, amplifying the blast radius beyond a single interaction.
To mitigate this risk, developers have implemented explicit detection logic for Mermaid blocks and applied HTML escaping to token content before wrapping it in the generated Mermaid structure. This ensures that special characters are converted into their safe entity equivalents, preventing them from being interpreted as executable code by the browser. Additionally, the patch moves the wrapping logic earlier in the page initialization process to ensure consistent protection across all Markdown instances. Organizations running affected versions should immediately upgrade to a patched release where these sanitization measures are active. Until an update is applied, administrators might consider restricting note editing permissions or implementing input validation at the application layer as temporary compensating controls.
This vulnerability aligns with Common Weakness Enumeration identifier CWE-79, which describes Improper Neutralization of Input During Web Page Generation commonly known as Cross-site Scripting. From a tactical perspective, it corresponds to MITRE ATT&CK technique T1059, specifically the execution of scripts within web browsers, and relates to persistence mechanisms where malicious content is stored for later activation. The failure to sanitize output before rendering in dynamic contexts highlights the critical need for robust input validation and output encoding strategies in modern web applications that support rich text or diagrammatic user inputs.