Evaluate the Stack, Not the Layer: Do Deterministic and LLM Gates for Agent Actions Fail Independently?

cs.AI

Submitted: 2026-10-05

Updated: 2026-10-05

Comments: 15 pages, 1 figure, 10 tables. Artifact (data, scripts, provenance): https://github.com/chenglin1112/evaluate-the-stack

Code: https://github.com/chenglin1112/evaluate-the-stack

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• AgentTrust: Runtime Safety Evaluation and Interception for AI Agent Tool Use
• Parallax: Why AI Agents That Think Must Never Act
• Layered LLM Defenses as an Ensemble: Access Tiers, Inference Cost, and the Measured Failure Correlation Between Defense Layers
• The Attacker Moves Second: Stronger Adaptive Attacks Bypass Defenses Against Llm Jailbreaks and Prompt Injections
• Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents
• AgentTrust: A Self-Improving Trust Layer for AI-Agent Actions
• What Can Be Enforced? A Theory of Certified Runtime Safety for Tool-Using Agents
• Nine Judges, Two Effective Votes: Correlated Errors Undermine LLM Evaluation Panels
• Great Models Think Alike and this Undermines AI Oversight
• Correlated Errors in Large Language Models
• Defeating Prompt Injections by Design
• Securing AI Agents with Information-Flow Control
• Progent: Securing AI Agents with Privilege Control
• Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations
• LlamaFirewall: An open source guardrail system for building secure AI agents
• AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents
• AgentTrap: Measuring Runtime Trust Failures in Third-Party Agent Skills
• LLMs Cannot Reliably Judge (Yet?): A Comprehensive Assessment on the Robustness of LLM-as-a-Judge
• Oversight Has a Capacity: Calibrating Agent Guards to a Subjective, Fatiguing Human

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems