CVE-2026-84809: Protection Mechanism Failure in Tencent AI-Infra-Guard

Tencent AI-Infra-Guard's skill-scan component fails to analyze compiled Python bytecode files because it explicitly excludes __pycache__ directories and files with .pyc, .pyo, and .pyd extensions from scanning. Attackers can exploit this by distributing skills containing benign Python source files alongside malicious compiled bytecode that executes on import. This causes the scanner to incorrectly report the skill as safe, enabling code execution when operators install the skill. The vulnerability affects versions >=0 <4.5.2, version 4.6.0, and versions <=0.2.1. The CVSS 4.0 vector indicates network attack vector, low attack complexity, no privileges required, user interaction required, and high impact on integrity.

An attacker can bypass the skill-scan component's protections by embedding malicious compiled Python bytecode files that are excluded from scanning. This allows execution of arbitrary code when the skill is installed, potentially compromising the system integrity. The vulnerability is rated high severity with a CVSS 4.0 score of 7.1.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, operators should be cautious when installing skills and consider additional manual inspection or alternative scanning methods for compiled bytecode files.