Inventree CVE-2026-35477

CVSS 5.5 with PR:H (high privilege required) significantly constrains real-world risk; this is not a remote unauthenticated attack. However, the vulnerability chain demonstrates a dangerous pattern: incomplete security fixes and inconsistent use of security controls create exploitability gaps. The attacker requires staff-level access and settings modification permissions, which limits the threat model to insider threats or accounts compromised with high privileges. SSVC analysis suggests this is remediation-triggered exposure: the incomplete fix to CVE-2026-27629 re-introduced the vulnerability in a form bypassing the intended sandbox. The tags confirm SSTI and RCE capability, elevating concern despite high privilege requirements. Organizations should prioritize patching based on their internal access controls; if staff accounts are widely distributed or poorly managed, risk escalates. The fact that versions 1.2.3-1.2.6 were shipped with a known-incomplete remediation is a process failure that warrants review of release practices.

A staff user with access to InvenTree's settings pages (specifically the PART_NAME_FORMAT configuration) crafts a Jinja2 template such as '{{ part.name }}{% if part.pk %}{{ __import__("os").system("id") }}{% endif %}' that includes conditional template injection. During validation, the sandboxed environment rejects the injection, but because the validator uses a dummy Part with pk=None, the conditional evaluates to False and the malicious code is never tested. The template is stored. When actual parts are rendered (pk has a real value), the production renderer uses the unsandboxed environment, the conditional evaluates to True, and arbitrary Python code executes with the privileges of the InvenTree application process. No public exploit code has been identified at the time of analysis; the attack requires intimate knowledge of the inconsistency between validator and renderer.