5ire CVE-2025-68669

Exploitation requires that a victim running 5ire version 0.15.2 or earlier renders attacker-controlled Markdown containing a Mermaid diagram; the markdown-it-mermaid plugin in useMarkdown.ts is initialized with securityLevel:'loose', which explicitly permits HTML tags inside Mermaid diagram nodes, so no special non-default setting is needed - the vulnerable configuration ships by default. The principal limiting factor is user interaction (UI:R): the malicious diagram content must actually be displayed/rendered by the victim (e.g., a malicious or prompt-injected model response or a shared conversation), and systems on a fixed release (post-0.15.2 with securityLevel:'strict') are not affected.

The NVD CVSS 3.1 score of 9.6 (Critical) is driven by a network attack vector (AV:N), low complexity (AC:L), no privileges (PR:N), and a scope change (S:C) with full C:H/I:H/A:H impact - consistent with a cross-site-scripting flaw (CWE-79) that escalates to code execution inside an Electron desktop app, where renderer-context HTML injection can reach the Node/OS layer. The one real brake on severity is UI:R: a victim must actually render attacker-supplied Markdown containing a Mermaid diagram, so this is not a wormable, fully unattended network exploit. A publicly available exploit is referenced in the GitHub Security Advisory (GHSA-5hpf-p8fw-j349), which raises practical risk, but there is NO CISA KEV listing and NO EPSS score in the provided data, so 'publicly available exploit code exists' must not be conflated with active exploitation - exploitation status is 'publicly available exploit code exists' only. Net assessment: a genuine high-priority issue for 5ire users because the delivery path (a malicious or prompt-injected AI response, or a shared conversation, rendering untrusted Mermaid/HTML) is realistic for an AI chat client, but overall real-world exposure is constrained by the product's small deployment footprint rather than by exploit difficulty. Note a data conflict: the description states 'not patched at time of publication,' yet commit 1fbe40d0 in the input changes securityLevel from 'loose' to 'strict,' indicating a fix now exists - verify current patched release with the vendor.

An attacker delivers a crafted Mermaid diagram containing embedded HTML/script to a 5ire user - for example via a malicious model response, a prompt-injection that causes the assistant to emit the payload, or a shared conversation. When the victim views the message, useMarkdown.ts renders the diagram with markdown-it-mermaid configured securityLevel:'loose', executing the injected HTML in the app's renderer and, in the Electron desktop context, leading to code execution. A public proof-of-concept is referenced in the GitHub advisory.