Agent security
Agent-specific security concerns and threats
Agent security frameworks and tools 102
- 🌟 Open Source AI Agent Security Infrastgithub.com
- Give your AI agent instant API loogithub.com
- How AI Agents Automate CVE Vulnerability Researchpraetorian.com
- SCAM — How safe is your AI agent?1password.github.io
- SuperClaw: Red-Team AI Agents Begithub.com
- Every API key you paste into an AI agent's input box hitsx.com
- AI Agent Lands PRs in Major OSS Projects, Targets Maintainersocket.dev
- Agentic SOC Platform: A pogithub.com
- An AI-powered agentic red team frameworgithub.com
- NeuralTrust - The Platform for AI and Agent Securityneuraltrust.ai
- Supply-chain risk of agentic AI - infecting infrastructures via skiblog.lukaszolejnik.com
- 'Signal' President and VP warn agentic AI is insecure, unreliable,coywolf.com
- Never Trust the Output: Data Pollution in AI Agents and MCPblog.slonser.info
- Runtime enforcement layer for AI agent agithub.com
- A curated list of 150+ pagithub.com
- Open-source AI agents for penetration testinggithub.com
- Block, Anthropic, and OpenAI Launch the Agentic AI Foundationblock.xyz
- Linux Foundation Announces the Formation of the Agentic AI Foundatilinuxfoundation.org
- 🔐Secure AI Agent Knowledge Retrieval - Introducing Security Filttechcommunity.microsoft.com
- How Code Execution Drives Key Risks in Agentic AI Systems - NVIDIAdeveloper.nvidia.com
- Autofix Bot: AI Agent purpose-built for code security - Autofix Botautofix.bot
- Introducing Aardvark: OpenAI's agentic security researcheropenai.com
- Leash by StrongDM — Security for AI Agentsleash.strongdm.ai
- securing ai agents.pptxdocs.google.com
- Agentic AI Red Teaming Playbookpillar.security
- AI agent for autonomous cybergithub.com
- Google DeepMind introduces new AI agent for code securitydeepmind.google
- Agentic AI Threat Modeling Framework: MAESTRO - CSAcloudsecurityalliance.org
- MAESTRO: Streamlining Agentic AI Security in IriusRiskiriusrisk.com
- Agentic Misalignment: How LLMs could be insider threatssimonwillison.net
- The lethal trifecta for AI agents: private data, untrusted content,simonwillison.net
- Unit 42 Develops Agentic AI Attack Frameworkpaloaltonetworks.com
- How to Hack AI Agents and Applicationsjosephthacker.com
- A security scanner for your LLM agegithub.com
- A Dynamic Environment to Evaluate Agithub.com
- Shinobi Security - Uncloud the cloud with your personal AI detectiveshinobi.security
- A curated list of GPT agents fgithub.com
- AARTS: An Open Standard for AI Agent Runtime Safetygithub.com
- Run NanoClaw in Docker Sandboxes with One Commandnanoclaw.dev
- Agent Governance Toolkit: Application-level security middleware for autonomous AI agentsgithub.com
- OpenShell - Safe, Private Runtime for Autonomous AI Agentsgithub.com
- Skills Security Index - Security Risk Analysis for Agentic AI Skill Definitionsindex.tego.security
- NemoClaw - Open Source Stack for Running OpenClaw Agents Safelygithub.com
- OneCLI - Secret Vault for AI Agentsgithub.com
- openclaw-a365 - Native Microsoft 365 Channel for OpenClawgithub.com
- Agent skills to help with Continuous Threagithub.com
- Fully-Autonomous AI Systems Are Discovering Vulns Todaymbgsec.com
- Streamlining Security Investigations with Agentsslack.engineering
- Agents testing framework made easygithub.com
- XBOW – Agents Built From Alloysxbow.com
- Charlemagne Labs - Safety is a Human Rightcharlemagnelabs.ai
- Clopus-Watcher: "intelligent" monitoringdenislavgavrilov.com
- AI Agent Benchmarks are Brokenddkang.substack.com
- Introducing tapes: transparent AI agent telemetryjohncodes.com
- Repo for "Adaptatgithub.com
- State of Agentic AI: Founder's edition - MMCmmc.vc
- Cursor "Open-Folder" Autorun Vulnerability Exposes Developers toasis.security
- Anthropic faces backlash to Claude 4 Opus behavior that contacts auventurebeat.com
- Wow. Opus 4.6 aggressively acquired and used authentication tokens and took unapprovedx.com
- The Agent Perimeter Fallacysecurityblueprints.io
- AARM - Autonomous Action Runtime Management, runtime security standard for AI agentsaarm.dev
- ADR - Agentic AI Detection and Response, deployed at Ubergithub.com
- Adrian - open-source runtime AI agent security monitorgithub.com
- Agent Baseline - six security outcomes for enterprise AI agentsagentbaseline.org
- agentseal - security toolkit for AI agents, scans skills and MCP configs, tests prompt injectiongithub.com
- AI agent shared responsibility model - Microsoft Azurelearn.microsoft.com
- Alex Volkov on NVIDIA announcing NemoClaw, a secure enterprise OpenClaw platform at GTCx.com
- ATR - Agent Threat Rules detection rule databaseagentthreatrule.org
- AVE - Behavioral Classification Standard for Agentic AIaveproject.org
- Clawdstrike - agentic AI EDR for developer workstations and agent swarmsgithub.com
- ClawSec - security skill suite for AI agentsgithub.com
- ClawVault - OpenClaw security vault controlling agent access to secrets and datagithub.com
- Cloudflare Mesh - secure private networking for users, nodes, and AI agentsblog.cloudflare.com
- DefenseClaw - security governance for OpenClaw and agentic AI runtimesgithub.com
- enforra - open source action governance SDK for AI agent tool callsgithub.com
- Geordie AI - AI agent governance, observability and control platformgeordie.ai
- h5i - secure, auditable browser for AI agentsgithub.com
- Introducing Prempti: Falco meets AI coding agentsfalco.org
- Introducing RAMPART and Clarity: Open Source Tools for Agent Development Safetymicrosoft.com
- iron-sensor - eBPF behavioral monitor for AI coding agentsgithub.com
- Kim Maida - The Agent Security Stack: Transport, Identity, Policy, Runtimex.com
- Microsoft Agent Governance Toolkit - OpenClaw sidecar deployment guidegithub.com
- microsoft/fides - Securing AI agents with information-flow controlgithub.com
- microsoft/fides tutorial - securing AI agents with information-flow controlgithub.com
- OpenClaw Security Platform - bring-your-own-security for AI agentszenitysec.github.io
- OpenClaw Security Platform - security layer for AI agentsgithub.com
- Pedro Franceschi on CrabTrap - an LLM-as-a-judge HTTP proxy to secure agents in productionx.com
- Prismor - AI agent security control planeprismor.dev
- SAM - Sovereign Agent Mesh, zero-trust P2P network for AI agentsgithub.com
- Securing Agents Across Perplexity's Client Endpoints with Numbatresearch.perplexity.ai
- sentra - external security provider for Microsoft Copilot Studiogithub.com
- SlowMist Agent Security Skill - security review framework for AI agentsgithub.com
- Straiker - the agentic AI security companystraiker.ai
- sycophant - zero-trust framework for autonomous AI agentsgithub.com
- tailcat - netcat-style encrypted tunnels from Tailscale for ad-hoc access without accounts or tailnetstailscale.com
- tirith - terminal security for developers and AI agentsgithub.com
- Trusted Remote Execution - policy-enforced scripts for AI agentsaws.amazon.com
- Uber open-sources ADR - Agentic Detection and Response framework and ADR-Benchx.com
- xaidr - runtime security for AI agentsgithub.com
- XBOW on how its autonomous offensive security agent architecture prevents OpenAI-Hugging Face style incidentsx.com
- Your AI Agent Is Indistinguishable From Malwareiron.sh
- Zero Trust for AI Agents - A security framework for deploying autonomous AI agents in the enterprisecdn.prod.website-files.com
Research 17
- AI Agent Smart Contract Exploit Generationarxiv.org
- Skill-Inject: Measuring Agent Vulnerability to Skill File Attacksarxiv.org
- Malicious Agent Skills in the Wild: A Large-Scale Security Empirical Studyarxiv.org
- SkillJect: Automating Stealthy Skill-Based Prompt Injection for Coding Agents with Trace-Driven Closed-Loop Refinementarxiv.org
- Agents of Chaosarxiv.org
- AI Agents Enable Adaptive Computer Wormscleverhans.io
- Browser AI Agents: The New 'Weakest Link' that Can Feed Your Credentials and Data to Attackerssqrx.com
- Emergent Cyber Behavior: When AI Agents Become Offensive Threat Actorsirregular.com
- endpoint-ai-agent-abuse - catalog of local AI agent abuse techniquesgithub.com
- GPT-5.5-Cyber built a zlib fuzzing lab in a dayblog.trailofbits.com
- Meta - Practical AI Agent Securityai.meta.com
- OpenClaw Agents Can Be Guilt-Tripped Into Self-Sabotagewired.com
- Patterns and problems in multiagent systems - Anthropic on coordination failures, collusion and sabotageanthropic.com
- Prelude shows Claude Code reading the macOS iMessage database despite EDRx.com
- Prismor - measuring secret exposure across 10,000 Claude Code agent sessionsx.com
- Watching Agents Work: A Behavioral Audit of Offensive-Security LLM Runsx.com
- When an Attacker Meets a Group of Agents: Navigating Amazon Bedrock's Multi-Agent Applicationsunit42.paloaltonetworks.com