Reflections on Trusting Trust, Revisited: Contaminating Self-Modifying AI Coding Agents with Poisoned Benchmarks

arXiv:2609.17817 · cs.CR, cs.AI · Submitted 2026-09-15 · Read on arXiv

cs.CR, cs.AI

Submitted: 2026-09-15

Updated: 2026-09-15

Code: https://github.com/ennyzzt/dgm

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

Terminology

• Benchmark — 217× in this paper · explained in 2 episodes
• Verification — 43× in this paper · explained in 2 episodes
• Scaffolding — 16× in this paper · explained in 3 episodes
• Agentic System — 7× in this paper · explained in 2 episodes
• AI Agents — 4× in this paper · explained in 2 episodes
• Accuracy — 4× in this paper · explained in 1 episode
• Self-Improving Coding Agent — 4× in this paper
• Agent Memory — 3× in this paper · explained in 1 episode
• Reward Hacking — 3× in this paper · explained in 8 episodes
• Robustness — 3× in this paper · explained in 5 episodes
• Agentic AI — 2× in this paper · explained in 1 episode
• Alignment — 2× in this paper · explained in 1 episode
• Core — 2× in this paper · explained in 2 episodes
• Multi-Agent System — 2× in this paper · explained in 6 episodes
• Prompt Engineering — 2× in this paper · explained in 4 episodes
• Redirect Server-Side Request Forgery — 2× in this paper
• Agent Harness — 1× in this paper · explained in 2 episodes
• Agent Interactions — 1× in this paper · explained in 1 episode
• Agentic AI systems — 1× in this paper · explained in 1 episode
• Agentic Coding — 1× in this paper · explained in 1 episode
• Agentic Reasoning — 1× in this paper · explained in 1 episode
• Analyzer — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 4 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Generalization — 1× in this paper · explained in 13 episodes
• Generative AI — 1× in this paper · explained in 2 episodes
• In International Conference on Learning Representations — 1× in this paper
• JSON Web Token — 1× in this paper
• LLM judge — 1× in this paper · explained in 2 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Reward Function — 1× in this paper · explained in 2 episodes
• Self-Evolving Agents — 1× in this paper · explained in 2 episodes
• Steering — 1× in this paper · explained in 2 episodes
• Trace — 1× in this paper · explained in 2 episodes

Sources

• A Systematic Review of Poisoning Attacks Against Large Language Models
• CaMeLs Can Use Computers Too: System-level Security for Computer Use Agents
• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
• The Red Queen G\"odel Machine: Co-Evolving Agents and Their Evaluators
• SWE-chat: Coding Agent Interactions From Real Users in the Wild
• MOSS: Self-Evolution through Source-Level Rewriting in Autonomous Agent Systems
• VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?
• Systems Security Foundations for Agentic Computing
• Prompt Infection: LLM-to-LLM Prompt Injection within Multi-Agent Systems
• Meta-Harness: End-to-End Optimization of Model Harnesses
• Trojan Hippo Bench: A Dynamic Benchmark for Persistent Memory Attacks and Defenses in LLM Agents
• Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies
• Alita-G: Self-Evolving Generative Agent for Agent Generation
• "Your AI, My Shell": Demystifying Prompt Injection Attacks on Agentic AI Coding Editors
• Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self-Evolution
• Prompt Injection Attacks on Agentic Coding Assistants: A Systematic Analysis of Vulnerabilities in Skills, Tools, and Protocol Ecosystems
• Supply-Chain Poisoning Attacks Against LLM Coding Agent Skill Ecosystems
• ceLLMate: Sandboxing Browser AI Agents
• Self-Improvements in Modern Agentic Systems: A Survey
• Agentic Much? Adoption of Coding Agents on GitHub

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs