YourMemory

YourMemory Tu IA tiene la memoria de un pez dorado. Ya no. Memoria persistente y auto-mejorable para agentes de IA, construida sobre la ciencia de cómo recuerdan los humanos. ▶ Prueba la demo interactiva en vivo · Sitio web · Evaluaciones comparativas El problema Cada mañana tu agente de IA te...

gpg_reaper

GPG Reaper TL;DR: Obtain/Steal/Restore GPG Private Keys from gpg-agent cache/memory This POC demonstrates method for obtaining GPG private keys from gpg-agent memory under Windows. Normally this should be possible only within 10 minutes time frame --default-cache-ttl value. Unfortunately...

Revoked but Still Authoritative: An Empirical Study of Revocation Enforcement in Agent-Memory Systems

Long-running language-model agents depend on persistent memory. Many agent-memory systems preserve history through soft revocation: a contradicted fact is marked invalid and retained rather than deleted. However, whether that mark is enforced at retrieval time is unexamined. In this paper, we...

CVE-2026-18993

The CVE-2026-18993 entry relates to NousResearch hermes-agent, specifically the Memory Toolset component, file hermes-agent/model_tools.py. The description indicates improper access controls in this functionality and an exploitable condition that can be triggered remotely. Affected versions inclu...

GHSA-G32J-MMXR-GFQ5 Flowise RCE via TypeORM DataSource

============================================================================= Security Advisory elttam Topic: Flowise RCE via TypeORM DataSource Module: FlowiseAI/Flowise Disclosed: 15-Apr-2026 Credits: Alex Brown Affects: FlowiseAI/Flowise 3.1.2 I. Background Flowise AI is an open-source, low-co...

PT-2026-67729

Name of the Vulnerable Software and Affected Versions Flowise versions prior to 3.1.3 Description Flowise record manager and agent memory nodes allow authenticated users to set arbitrary TypeORM DataSource options via the additionalConfig input. This occurs in the following components: -...

MemSecBench: Tracking Agent Memory Poisoning from Persistence to Consequence and Repair

Memory systems allow agents to retain and reuse information from past interactions, but they can also let malicious content persist. A malicious instruction crafted by an attacker may be stored in long-term memory, recalled much later, and quietly shape a real action. Recent benchmarks increasing...

CVE-2026-5998: Improper Limitation of a Pathname to a Restricted Directory

A flaw has been found in zhayujie chatgpt-on-wechat CowAgent up to 2.0.4. This affects the function dispatch of the file agent/memory/service.py of the component API Memory Content Endpoint. This manipulation of the argument filename causes path traversal. The attack can be initiated remotely. Th...

PT-2026-30464

🚨 LIVE HIJACK ALERT — CVE-2026-77777. CVSS 8.9. attacker feeds poisoned task output to CrewAI agent memory. future tasks execute attacker instructions as if they were original goals. investigating. 🧵...

Poison Once, Exploit Forever: Environment-Injected Memory Poisoning Attacks on Web Agents

Memory makes LLM-based web agents personalized, powerful, yet exploitable. By storing past interactions to personalize future tasks, agents inadvertently create a persistent attack surface that spans websites and sessions. While existing security research on memory assumes attackers can directly...

SUSE-SU-2026:0215-1 Security update for gpg2

This update for gpg2 fixes the following issues: - CVE-2025-68973: Fix possible memory corruption in the armor parser gpg.fail/memcpybsc1255715. - Avoid potential downgrade to SHA1 in 3rd party key signatures gpg.fail/sha1 bsc1256246. - Error out on unverified output for non-detached signatures...

CVE-2022-28946

An issue in the component ast/parser.go of Open Policy Agent v0.39.0 causes the application to incorrectly interpret every expression, causing a Denial of Service DoS via triggering out-of-range memory access...

A Practical Memory Injection Attack against LLM Agents

In this whitepaper, the authors propose a novel Memory INJection Attack, MINJA, that enables the injection of malicious records into the memory bank by only interacting with the agent via queries and output observations...

agent-memory-layer (>=0.1.0, <=0.1.1), hana-ai (>=1.0.25122300, <=1.1.26072400) +2 more potentially affected by CVE-2023-25617 via sap-ai-sdk-base (>=3.1.2, <=3.4.0)

sap-ai-sdk-base PYPI version =3.1.2, =0.1.0, =1.0.25122300, =3.0.3, =5.3.4, =7.4.1 Source cves: CVE-2023-25617 Source advisory: SNYK:PYTHON-SAPAISDKBASE-17670581...

CVE-2022-33751

CA Automic Automation 12.2 and 12.3 contain an insecure memory handling vulnerability in the Automic agent that could allow a remote attacker to potentially access sensitive data...

Security Bulletin: IBM WebSphere MQ Heap storage leak on error path could lead to denial of service (CVE-2016-0260)

Summary Heap storage allocated on an error path is not deallocated by queue manager agents, a malicious user could use an application to repeatedly cause the queue manager to execute this error path and eventually exhaust storage for the agent, thereby causing a denial of service. Vulnerability...