Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback

• URL: http://arxiv.org/abs/2605.17453v1
• Date: Sun, 17 May 2026 13:51:34 GMT
• Title: Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback
• Abstract summary: We study a different failure mode, cognitive poisoning, in which a malicious tool behaves plausibly during exploration.<n>We present VISTA-Guard, a backbone-agnostic framework for final-action risk scoring.
• Score: 38.251599309409
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Tool-using LLM agents increasingly rely on external tools to make consequential decisions, yet most existing agent-security benchmarks and defenses implicitly assume that tool feedback is trustworthy once a tool has been selected. We study a different failure mode, cognitive poisoning, in which a malicious tool behaves plausibly during exploration, accumulates trust through benign-looking feedback, and becomes harmful only when hidden state conditions align with the final executable action. To study this setting, we construct TRUST-Bench, a task-conditioned benchmark of 1,970 hidden-trigger tool-compromise episodes with matched safe controls, introduce an asymmetric penalty metric, GuardedJoint, to better reflect real deployment risk, and present VISTA-Guard, a backbone-agnostic framework for final-action risk scoring. The core idea is to abstract multi-step tool interaction into structured environment variables that encode trust-formation dynamics and then score the risk of the final executable action from this trajectory-conditioned representation. Experiments show that prompt-centric heuristics, scalarized features, and zero-shot judges fail in this regime, whereas trajectory-aware final-action scoring yields strong in-domain discrimination and remains effective under balanced out-of-distribution transfer. Under GuardedJoint, VISTA-Guard reaches $84.2$ in-domain and $56.9$ on balanced out-of-distribution evaluation, while methods that optimize only one side of the safety--utility tradeoff collapse to zero. These findings support a broader view of agent security in black-box tool ecosystems: the decisive defense target is not local prompt text or tool descriptors alone, but the way trust is formed across the interaction trajectory and committed through the final action.

Related papers

• BLINDSPOT: A Benchmark for Safety and Refusal Calibration in Long-Horizon Tool-Using Agents [20.217145643468598]
  We introduce Blindspot, a benchmark for trajectory-level safety calibration of long-horizon tool-using agents.<n>Blindspot evaluates complete user-agent-environment trajectories through adaptive adversarial interaction, stateful tool execution, and execution-grounded adjudication.<n>We evaluate 13 proprietary and open-weight LLMs using eight metrics covering unsafe completion, appropriate refusal, benign utility, over-refusal, repeated-run robustness, and post-refusal failure.
  arXiv  Detail & Related papers  (2026-09-14T20:12:27Z)
• HazardAuditor: From Executable Threats to Safer Computer-Use Agents [50.06351661912873]
  Existing guard models target static prompts and responses.<n>Existing safety platforms produce evaluation verdicts rather than the normalized supervision a guard model needs to learn.<n>We introduce HazardAuditor, an execution-grounded framework that closes both gaps.
  arXiv  Detail & Related papers  (2026-09-14T07:09:33Z)
• ROPE: Routed Origin Policy Enforcement against Indirect Prompt Injection [61.4148196085256]
  We present ROPE (Routed Origin Policy Enforcement), which is anchored in a structural notion of trust.<n>Our approach admits two provable guarantees: 1) at every step of a trajectory, no value whose only origin is attacker-writable content reaches an origin-guarded parameter, and 2) no rewording of an injection changes an admission decision.
  arXiv  Detail & Related papers  (2026-08-27T01:22:14Z)
• SPADE-Bench: Evaluating Spontaneous Strategic Deception in Agents via Plan-Action Divergence [30.99083906950254]
  We introduce SPADE-Bench, a benchmark designed to evaluate spontaneous plan-action divergence.<n>Unlike prior deception benchmarks, SPADE-Bench simultaneously integrates actual tool execution and controlled pressure scenarios.<n> Experiments confirm that agent deception is a genuine and pressing issue in tool-use contexts.
  arXiv  Detail & Related papers  (2026-06-01T15:28:34Z)
• On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment [54.30690671490447]
  Existing safety-alignment signals are largely response-level or off-policy.<n>FATE transforms verifier-scored failures into repair supervision without expert demonstrations.<n>FATE reduces attack success rate by 33.5%, harmful compliance by 82.6%, and improves external trajectory-safety diagnosis by 6.5%.
  arXiv  Detail & Related papers  (2026-05-12T09:56:28Z)
• The Causal Impact of Tool Affordance on Safety Alignment in LLM Agents [0.3823356975862005]
  This work empirically examines how executable tool affordance alters safety alignment in large language models (LLMs)<n>Experiments are conducted in a deterministic financial transaction environment with binary safety constraints across 1,500 scenarios.<n>Both evaluated models maintain perfect compliance in text-only settings, yet exhibit sharp increases in violations after tool access is introduced.
  arXiv  Detail & Related papers  (2026-03-19T23:34:46Z)
• Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents [68.20752678837377]
  We propose a principled taxonomy that transforms single-turn harmful tasks into multi-turn attack sequences.<n>Using this taxonomy, we construct MT-AgentRisk, the first benchmark to evaluate multi-turn tool-using agent safety.<n>We propose ToolShield, a training-free, tool-agnostic, self-exploration defense.
  arXiv  Detail & Related papers  (2026-02-13T18:38:18Z)
• DRAFT: Task Decoupled Latent Reasoning for Agent Safety [59.46137757545185]
  We propose DRAFT (Task Decoupled Latent Reasoning for Agent Safety), a latent reasoning framework that decouples safety judgment into two trainable stages.<n>An Extractor distills the full trajectory into a compact continuous latent draft, and a Reasoner that jointly attends to the draft and the original trajectory to predict safety.<n>DRAFT consistently outperforms strong baselines, improving accuracy from 63.27% (LoRA) to 91.18% averaged over benchmarks.
  arXiv  Detail & Related papers  (2026-02-11T07:45:14Z)
• ReasAlign: Reasoning Enhanced Safety Alignment against Prompt Injection Attack [52.17935054046577]
  We present ReasAlign, a model-level solution to improve safety alignment against indirect prompt injection attacks.<n>ReasAlign incorporates structured reasoning steps to analyze user queries, detect conflicting instructions, and preserve the continuity of the user's intended tasks.
  arXiv  Detail & Related papers  (2026-01-15T08:23:38Z)
• VIGIL: Defending LLM Agents Against Tool Stream Injection via Verify-Before-Commit [44.24310459184061]
  LLM agents operating in open environments face escalating risks from indirect prompt injection.<n>We propose textbfVIGIL, a framework that shifts the paradigm from restrictive isolation to a verify-before-commit protocol.
  arXiv  Detail & Related papers  (2026-01-09T12:19:49Z)
• Quantifying Distributional Robustness of Agentic Tool-Selection [8.457056023589951]
  We introduce ToolCert, the first statistical framework that formally certifies tool selection robustness.<n>We show that ToolCert produces a high-confidence lower bound on accuracy, formally quantifying the agent's worst-case performance.<n>Our evaluation with ToolCert uncovers the severe fragility: under attacks injecting deceptive tools or saturating retrieval, the certified accuracy bound drops near zero.
  arXiv  Detail & Related papers  (2025-10-05T01:50:34Z)
• STAC: When Innocent Tools Form Dangerous Chains to Jailbreak LLM Agents [38.755035623707656]
  This paper introduces Sequential Tool Attack Chaining (STAC), a novel multi-turn attack framework that exploits agent tool use.<n>We apply our framework to automatically generate and evaluate 483 STAC cases, featuring 1,352 sets of user-agent-environment interactions.<n>Our evaluations show that state-of-the-art LLM agents, including GPT-4.1, are highly vulnerable to STAC, with attack success rates (ASR) exceeding 90% in most cases.
  arXiv  Detail & Related papers  (2025-09-30T00:31:44Z)
• Dissecting Adversarial Robustness of Multimodal LM Agents [70.2077308846307]
  We manually create 200 targeted adversarial tasks and evaluation scripts in a realistic threat model on top of VisualWebArena.<n>We find that we can successfully break latest agents that use black-box frontier LMs, including those that perform reflection and tree search.<n>We also use ARE to rigorously evaluate how the robustness changes as new components are added.
  arXiv  Detail & Related papers  (2024-06-18T17:32:48Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.