Breaking the Protocol: Security Analysis of the Model Context Protocol Specification and Prompt Injection Vulnerabilities in Tool-Integrated LLM Agents

servers to claim arbitrary permissions, (2) bidirectional sampling without origin authentication enabling server-side prompt injection, and (3) implicit trust propagation in multi-server configurations. We implement \textsc{MCPBench

ASTRIDE: A Security Threat Modeling Platform for Agentic-AI Applications

large language models (LLMs). However, these systems introduce novel and evolving security challenges, including prompt injection attacks, context poisoning, model manipulation, and opaque agent-to-agent communication, that

Countermind: A Multi-Layered Security Architecture for Large Language Models

Large Language Model (LLM) applications is fundamentally challenged by "form-first" attacks like prompt injection and jailbreaking, where malicious instructions are embedded within user inputs. Conventional defenses, which rely

/pinclaw/send. Unauthenticated attackers reaching port 18790, which binds all interfaces by default, can inject blind prompts into the user's main OpenClaw agent session as user instructions

Analyzing the Narration Gap in LLM-Solver Loops

loop as a verified decision procedure. We further evaluate five open-sourced models under prompt injection, and we find certificate gating makes the solver verdict sound, while an adversary

Minimal Prompt Perturbations Lead to Code Vulnerabilities: Prompt Fragility and Hidden-State Signals in Coding LLMs

These results show that the threat model for LLM-assisted coding extends beyond prompt injection to ordinary prompt variation, and indicate that input-handling flaws can be caught before generation

Owner-Harm: A Missing Threat Model for AI Agent Safety

criminal harm) yet only 14.8% (4/27; 95% CI: 5.9%-32.5%) on AgentDojo injection tasks (prompt-injection-mediated owner harm). A controlled generic-LLM baseline shows the gap is not inherent

PraisonAI is a multi-agent teams system. Prior to 4.5.128

Detection of adversarial intent in Human-AI teams using LLMs

useful, it also exposes them to a broad range of attacks, including data poisoning, prompt injection, and even prompt engineering. Through these attack vectors, malicious actors can manipulate

Scope-Gap Injection: Exploiting Evaluator Channel Blind Spots in the OpenClaw Agent Security Competition

This paper introduces Scope-Gap Injection (SGI), an indirect prompt injection method for multi-stage LLM-agent safety pipelines. In the IJCAI 2026 OpenClaw Agent Security Attack Competition, each attack

TrustShiftProbe: Characterizing, Benchmarking, and Defending Staged Trust Attacks on MCP Servers

filters. Crucially, TrustShift originates in the server-controlled tool channel, not user prompts (unlike indirect prompt injection) or the transport (unlike man-in-the-middle): the adversary is the trusted

Mitigating Visual Degradation in MLLMs via Spatial-Spectral Visual Anchor Learning

propose Spatial-Spectral Visual Anchor Learning (SSVAL). The core of SSVAL is Visual Anchor Prompt Injection (VAPI), which introduces prompts that absorb rich knowledge from external VFMs during training, enabling

Salience Induction against Multi-Hop RAG Agents: Threat and Defense

facts across documents. Existing defenses focus on content poisoning, which injects false facts, and prompt injection, which embeds directives. We identify a third attack surface: the salience channel, through which

Beyond Wireless Security: Covert Communications in Large Language Model-enabled Edge Networks

computations make LLMENs susceptible to various security threats, such as eavesdropping, jamming, prompt poisoning, and prompt injection attacks. Since existing countermeasures against these attacks often incur prohibitive overhead, developing holistic

Fortytwo: Swarm Inference with Peer-Ranked Consensus

evaluation indicates higher accuracy and strong resilience to adversarial and noisy free-form prompting (e.g., prompt-injection degradation of only 0.12% versus 6.20% for a monolithic single-model baseline), while

OpenClaw versions 2026.2.13 through 2026.3.24 contain an ANSI escape sequence injection vulnerability in approval prompts that allows attackers to spoof terminal output. Untrusted tool metadata can carry ANSI control sequences

Agentic Cloud Decoys: A Deception-Driven Framework for Autonomous Intrusion Investigation

attacker chosen values providers record verbatim, which makes any log to prompt path an indirect prompt injection channel that a decoy widens rather than narrows. We address the first

SafeClawBench: Separating Semantic, Audit-Evidence, and Sandbox Harm in Tool-Using LLM Agents

agent security with 600 controlled adversarial tasks across six attack families: direct and indirect prompt injection, tool-return injection, memory poisoning, memory extraction, and ambiguity-driven unsafe inference. SafeClawBench reports

MobileWorldSafety: Benchmarking GUI Agent Safety Against Environmental Injection Attacks in Android Apps

untrusted environmental content, they are highly vulnerable to environmental injection attacks, which include indirect prompt injections and adversarial instructions. Such attacks can manipulate the behavior of agents without user awareness

CIBER: A Comprehensive Benchmark for Security Evaluation of Code Interpreter Agents

vulnerability of code interpreter agents against four major types of adversarial attacks: Direct/Indirect Prompt Injection, Memory Poisoning, and Prompt-based Backdoor. We evaluate six foundation models across two representative code