Two agent-skill security scanners were both blind to __pycache__, letting a malicious skill ship a clean verdict

NVD·high signalXBlueskyLinkedInCopy link

CVE-2026-84809 and CVE-2026-84811, both published 2026-09-02 at CVSS 7.1, hit Tencent's AI-Infra-Guard skill-scan and agentverus-scanner with the identical flaw: each hardcodes __pycache__ directories and .pyc/.pyo/.pyd extensions into skip lists, so an attacker ships benign Python source next to malicious compiled bytecode that executes on import. agentverus-scanner returns a CERTIFIED verdict with high trust scores in both static and semantic modes while the payload runs. If you gate skill installs on either scanner, the gate does not see the file that runs.