The vulnerability is a server‑side template injection in the LMFDB application’s render_knowl function that allows an unauthenticated user to inject arbitrary Jinja expressions through the /knowledge/render/<ID> endpoint. By supplying crafted query-string parameters or content data, an attacker can reach Python built‑ins and execute operating‑system commands with the web application’s account privileges. This flaw enables full remote code execution on the host running LMFDB.

The flaw affects all releases of LMFDB before commit 01829bd. The product is the LMFDB web‑application, and any installation that has not applied the patch that introduces the commit 01829bd fix is vulnerable. There is no version suffix given beyond the commit identifier, so any instance that has not incorporated this change is at risk.

The CVSS score is 9.3, indicating a critical risk. The EPSS score is not available, so the likelihood of exploitation cannot be quantified via that metric, but the vulnerability is openly documented and the flaw was publicly disclosed. The issue is not listed in CISA’s KEV catalog, yet the ability to trigger OS command execution without authentication makes it a high‑priority target for attackers. The attack vector is likely a remote HTTP request to the vulnerable endpoint, and no special privileges or network access are required beyond the ability to reach the web server.