InjecMEM: Memory Injection Attack on LLM Agent Memory Systems

Memory is becoming a default subsystem in deployed LLM agents to provide persistent personalization and continuity. This naturally prompts a question: will memory system introduce new vulnerabilities into agents? Thus we propose InjecMEM, a novel memory injection attack paradigm that requires onl...

PT-2026-80890

Improper Neutralization of Special Elements used in an OS Command in the package manager component of Black Duck blackduck-c-cpp before 3.0.7 allows an actor able to create a file within the scanned build directory to execute operating system commands as the account running the scan. Filesystem...

PT-2026-80986

The Continue CLI applies an incomplete denylist as its only barrier to destructive shell commands when running unattended. In headless mode and auto mode the default policy in extensions/cli/src/permissions/defaultPolicies.ts grants the Bash tool the allow permission, and permissionChecker.ts...

PT-2026-80949

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the jsonstatus function. The vulnerability is caused by insufficient filtering of the usescript, usefile, and option fields before command execution. A remote attacker can trigger this vulnerability via crafted input...

PT-2026-80941

Multiple DrayTek VigorAP models contain a command injection vulnerability in the setcamset function. The vulnerability is caused by insufficient filtering of the selectSlaves field before command execution. A remote attacker can trigger this vulnerability via crafted input to execute arbitrary...

PT-2026-79838

Name of the Vulnerable Software and Affected Versions DJI Neo versions prior to 01.00.0400 DJI Neo 2 versions prior to 01.00.0500 DJI Flip versions prior to 01.00.1200 DJI Air 3 versions prior to 01.00.1600 DJI Air 3S versions prior to 01.00.1400 DJI Avata 2 versions prior to 01.00.0400 DJI Avata...

PT-2026-80977

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the setDevNet function. The vulnerability is caused by insufficient filtering of the username and password fields before command execution. A remote attacker can trigger this vulnerability via crafted input to execut...

PT-2026-80979

An authenticated command injection vulnerability in TP-Link Archer BE800 V1 allows an attacker with administrative access to execute arbitrary system commands with root privileges by injecting shell metacharacters via a VPN connection. Successful exploitation may enable persistent backdoors,...

PT-2026-80958

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the getVid function. The vulnerability is caused by insufficient filtering of the username and password fields before command execution. A remote attacker can trigger this vulnerability via crafted input to execute...

PT-2026-80951

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the pingtrace function. The vulnerability is caused by insufficient validation of the host field before command execution. A remote attacker can trigger this vulnerability via crafted input to execute arbitrary...

PT-2026-80961

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the rebDevice function. The vulnerability is caused by insufficient filtering of the username and password fields before command execution. A remote attacker can trigger this vulnerability via crafted input to execut...

PT-2026-80955

Multiple DrayTek VigorSwitch models contain a pre-authentication command injection vulnerability in the setget.cgi interface. The vulnerability is caused by insufficient filtering of the pass field before command execution. A remote attacker can trigger this vulnerability via crafted input to...

PT-2026-80957

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the auth set function. The vulnerability is caused by insufficient filtering of the username and password fields before command execution. A remote attacker can trigger this vulnerability via crafted input to execute...

PT-2026-80993

A stored OS command injection vulnerability exists in the parent-control module of TP-Link Archer BE3600 V1. An authenticated adjacent attacker with administrative access may store a crafted profile name containing shell metacharacters, which is later processed unsafely during daily cloud report...

PT-2026-80960

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the setDevice function. The vulnerability is caused by insufficient sanitization of the username, password, and location fields before command execution. A remote attacker can trigger this vulnerability via crafted...

PT-2026-80963

Multiple DrayTek VigorSwitch models contain a command injection vulnerability in the setDevProto function. The vulnerability is caused by insufficient filtering of the username and password fields before command execution. A remote attacker can trigger this vulnerability via crafted input to...

PT-2026-81115

Commit: f855e54d5b7bc1c91b977574a03b91eff6b86bb6 Component: mcp-servers/python/python sandbox server/src/python sandbox server/server fastmcp.py Vulnerability RestrictedPython's sandbox in ContextForge's python sandbox server sub-project allows arbitrary Python execution via three compounding...

PT-2026-80947

Multiple DrayTek VigorAP models contain a command injection vulnerability in the upload settings.cgi interface. The vulnerability is caused by insufficient filtering before the restorekey field is concatenated into a shell command. A remote attacker can trigger this vulnerability via crafted inpu...

PT-2026-80800

Firmware in KAON PG5298A and PG5298B routers allow an authenticated user to send crafted JSON-RPC requests and perform operations not possible via GUI, e.g. system file read or command execution. This vulnerability has been fixed in firmware version: 3.0.82 for PG5298A and 4.0.82 for PG5298B...

PT-2026-80940

Multiple DrayTek VigorAP models contain a command injection vulnerability in the setLan function. The vulnerability is caused by insufficient validation of the lanIp and lanNetmask fields before command execution. A remote attacker can trigger this vulnerability via crafted input to execute...