Agent-ValueBench is the first dedicated benchmark for agent values, showing they diverge from LLM values, form a homogeneous 'Value Tide' across models, and bend under harnesses and skill steering.
hub
Don’t let the claw grip your hand: A security analysis and defense framework for OpenClaw
18 Pith papers cite this work. Polarity classification is still indexing.
18
Pith papers citing it
hub tools
citation-role summary
background 2
baseline 1
citation-polarity summary
years
2026 18representative citing papers
AgentCanary introduces an Entry × Impact risk taxonomy, high-fidelity real tool environments with persistent state, and multi-dimensional trajectory evaluation to assess AI agent security across models and attacks.
The study identifies four memory write channels and nine structural vulnerabilities in LLM agents, proposes a taxonomy of six attack classes, introduces MPBench, and finds that aggressive memory use increases exploitability while existing defenses fail.
AgentHazard benchmark shows computer-use agents remain highly vulnerable, with attack success rates reaching 73.63% on models like Qwen3-Coder powering Claude Code.
SeClaw provides spec-driven synthesis of security tasks and an execution-based docker testbed for evaluating unsafe behaviors in autonomous LLM agents.
A new multi-turn security benchmark shows OpenClaw agents across ten LLMs trigger harmful actions in 28–53% of adversarial cases, and fragmented or file-hidden attacks roughly double baseline risk rates.
LivePI benchmark reports indirect prompt injection success rates of 10.7-29.6% across five models on seven input surfaces and shows a two-layer defense blocking all malicious completions while preserving utility.
All six evaluated OpenClaw agent frameworks exhibit substantial security vulnerabilities, with reconnaissance behaviors as the most common weakness and agent systems proving significantly riskier than isolated backbone models.
Claw AI agents' heartbeat background execution shares memory context with user sessions, allowing ordinary social misinformation to silently pollute long-term memory and shape behavior at rates up to 76% across sessions.
PolicyAlign aligns LLMs to natural-language safety policies by synthesizing violating instructions and performing on-policy self-distillation with policy-sensitive filtering, improving safety without high-quality supervision data.
Introduces a compositional governance framework defining delegation types, resource scope attenuation, and an overlay operator for agentic AI authorization policies.
BraveGuard trains guard models on realistic agent trajectories derived from open-world threats, raising detection accuracy on AgentHazard from 38.79% to 82.38%.
LLM agent security is reframed as an agent-human interaction issue, supported by a survey showing industry preference for human-centric mechanisms over academic favorites and proposing a new research agenda.
Generative AI advertising is reframed as a problem of trustworthy commercial intervention on the generative process, with a taxonomy of influence tiers from product mentions to long-term preference shaping.
A TEE-backed architecture isolates security-critical decisions in self-hosted AI agents to prevent host-level abuse from malicious inputs while maintaining allowed functionality.
AgentOpt introduces a framework-agnostic package that uses algorithms like UCB-E to find cost-effective model assignments in multi-step LLM agent pipelines, cutting evaluation budgets by 62-76% while maintaining near-optimal accuracy on benchmarks.
This work categorizes seven risks of OpenClaw for non-technical users, provides plain-language mitigations, and supplies a companion Skill to automate security configurations.
A survey that categorizes threats to OpenClaw agents including skill poisoning and cognitive manipulation and reviews defense mechanisms.
citing papers explorer
-
Agent-ValueBench: A Comprehensive Benchmark for Evaluating Agent Values
Agent-ValueBench is the first dedicated benchmark for agent values, showing they diverge from LLM values, form a homogeneous 'Value Tide' across models, and bend under harnesses and skill steering.
-
AgentCanary: A Security Evaluation Framework for Autonomous AI Agents in Real Executable Environments
AgentCanary introduces an Entry × Impact risk taxonomy, high-fidelity real tool environments with persistent state, and multi-dimensional trajectory evaluation to assess AI agent security across models and attacks.
-
From Untrusted Input to Trusted Memory: A Systematic Study of Memory Poisoning Attacks in LLM Agents
The study identifies four memory write channels and nine structural vulnerabilities in LLM agents, proposes a taxonomy of six attack classes, introduces MPBench, and finds that aggressive memory use increases exploitability while existing defenses fail.
-
AgentHazard: A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents
AgentHazard benchmark shows computer-use agents remain highly vulnerable, with attack success rates reaching 73.63% on models like Qwen3-Coder powering Claude Code.
-
SeClaw: Spec-Driven Security Task Synthesis for Evaluating Autonomous Agents
SeClaw provides spec-driven synthesis of security tasks and an execution-based docker testbed for evaluating unsafe behaviors in autonomous LLM agents.
-
ASEval: Automated Trajectory-Level Security Testing for Autonomous Agents
A new multi-turn security benchmark shows OpenClaw agents across ten LLMs trigger harmful actions in 28–53% of adversarial cases, and fragmented or file-hidden attacks roughly double baseline risk rates.
-
LivePI: More Realistic Benchmarking of Agents Against Indirect Prompt Injection
LivePI benchmark reports indirect prompt injection success rates of 10.7-29.6% across five models on seven input surfaces and shows a two-layer defense blocking all malicious completions while preserving utility.
-
A Systematic Security Evaluation of OpenClaw and Its Variants
All six evaluated OpenClaw agent frameworks exhibit substantial security vulnerabilities, with reconnaissance behaviors as the most common weakness and agent systems proving significantly riskier than isolated backbone models.
-
Mind Your HEARTBEAT! Claw Background Execution Inherently Enables Silent Memory Pollution
Claw AI agents' heartbeat background execution shares memory context with user sessions, allowing ordinary social misinformation to silently pollute long-term memory and shape behavior at rates up to 76% across sessions.
-
PolicyAlign: Direct Policy-Based Safety Alignment for Large Language Models
PolicyAlign aligns LLMs to natural-language safety policies by synthesizing violating instructions and performing on-policy self-distillation with policy-sensitive filtering, improving safety without high-quality supervision data.
-
Overlaying Governance: A Compositional Authorization Framework for Delegation and Scope in Agentic AI
Introduces a compositional governance framework defining delegation types, resource scope attenuation, and an overlay operator for agentic AI authorization policies.
-
BraveGuard: From Open-World Threats to Safer Computer-Use Agents
BraveGuard trains guard models on realistic agent trajectories derived from open-world threats, raising detection accuracy on AgentHazard from 38.79% to 82.38%.
-
Reframing LLM Agent Security as an Agent-Human Interaction Problem
LLM agent security is reframed as an agent-human interaction issue, supported by a survey showing industry preference for human-centric mechanisms over academic favorites and proposing a new research agenda.
-
Generative AI Advertising as a Problem of Trustworthy Commercial Intervention
Generative AI advertising is reframed as a problem of trustworthy commercial intervention on the generative process, with a taxonomy of influence tiers from product mentions to long-term preference shaping.
-
Constraining Host-Level Abuse in Self-Hosted Computer-Use Agents via TEE-Backed Isolation
A TEE-backed architecture isolates security-critical decisions in self-hosted AI agents to prevent host-level abuse from malicious inputs while maintaining allowed functionality.
-
AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agent
AgentOpt introduces a framework-agnostic package that uses algorithms like UCB-E to find cost-effective model assignments in multi-step LLM agent pipelines, cutting evaluation budgets by 62-76% while maintaining near-optimal accuracy on benchmarks.
-
Understanding and mitigating the risks of OpenClaw for non-technical users: A practical guide with Skill
This work categorizes seven risks of OpenClaw for non-technical users, provides plain-language mitigations, and supplies a companion Skill to automate security configurations.
-
Security of OpenClaw Agents: Fundamentals, Attacks, and Countermeasures
A survey that categorizes threats to OpenClaw agents including skill poisoning and cognitive manipulation and reviews defense mechanisms.