When Tool Outputs Become Commands: Separating Action Induction from Runtime Authorization in Tool-Augmented LLM Agents

• URL: http://arxiv.org/abs/2608.27146v1
• Date: Thu, 27 Aug 2026 13:59:04 GMT
• Title: When Tool Outputs Become Commands: Separating Action Induction from Runtime Authorization in Tool-Augmented LLM Agents
• Abstract summary: We propose SARA, which treats action induction and execution authorization as distinct runtime roles.<n>Across AgentDojo and AgentDyn, SARA limits ASR to no more than (0.63%) across four primary evaluation settings.
• Score: 11.018574747618173
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Tool-augmented LLM agents must rely on untrusted runtime Observations to complete open-ended tasks; however, when tool outputs no longer merely provide data but begin to specify concrete actions, they effectively become ``commands'' that can drive real-world side effects beyond user intent. We argue that this risk arises from conflating action induction with execution authorization. To address this distinction, we propose SARA, which treats action induction and execution authorization as distinct runtime roles and separates action provenance from execution authority. On the Observation side, a context-isolated Action Probe exposes action-inducing semantics and persistently records action-origin provenance across steps as a review signal; on the execution side, actual tool calls are authorized only against the user objective and audited evidence from authorized successful executions, while satisfying goal, execution-chain, and argument-level support. To preserve this separation across multi-step execution, SARA applies No-History-Promotion to prevent historical recurrence from laundering action origins into execution authority. Across AgentDojo and AgentDyn, SARA limits ASR to no more than \(0.63\%\) across four primary evaluation settings while maintaining competitive task utility, and consistently reduces ASR across additional Agent backbones.

Related papers

• ActGuard: Pre-execution Action Auditing against Indirect Prompt Injection in LLM Agents [11.495797074855703]
  We propose ActGuard, a pre-execution action auditing framework.<n>Rather than judging whether external content is inherently suspicious, ActGuard assesses whether it causes the current action to deviate from a reasonable expectation.<n>A verifier then examines the localized evidence, masks only spans confirmed as malicious, and regenerates the action from the sanitized context.
  arXiv  Detail & Related papers  (2026-09-14T03:47:05Z)
• From Intent to Execution Grant: An Execution-Boundary Conformance Profile for High-Risk AI Actions [9.27474112226667]
  EBL-Core decides whether one canonical, fully materialized AI-generated candidate may receive action-scoped execution authority under explicit conditions.<n>It binds, policy non-weakening, evidence handling, deterministic adjudication, verification, and grant lifecycle behavior.<n>Across 100 trials, 32 concurrent Redemption attempts yielded exactly one successful Redemption and protected test effect per trial.
  arXiv  Detail & Related papers  (2026-09-10T14:21:45Z)
• From Event Logs to Governed Action: A BlueSky Agenda for Agentic Process Mining [72.81644384922838]
  We propose event-to-action process mining: a process-mining agenda for transforming heterogeneous operational event data into governed action.<n>We argue that the community needs four mineable artifacts: event-object representations, action evidence packages, governance contracts, and benchmarks where act, defer, ask, and refuse are all valid outputs.
  arXiv  Detail & Related papers  (2026-09-07T21:02:23Z)
• 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)
• AgentAbstain: Do LLM Agents Know When Not to Act? [19.223843408018997]
  Existing evaluations mostly focus on task success rather than whether agents know when to abstain.<n>We present the first systematic evaluation framework for agentic abstention: the calibrated ability of tool-using LLM agents to recognize when not to act.<n>AgentAbstain is a paired-task benchmark built on an agent-native taxonomy of 8 abstention scenarios across pre-execution reasoning and runtime discovery.
  arXiv  Detail & Related papers  (2026-07-11T00:57:36Z)
• Context-to-Execution Integrity for LLM Agents [0.0]
  Context-to-Execution Integrity ( CXI) is an execution-boundary system for this setting.<n>We evaluate CXI on open-weight field-projection runs, AgentDojo live episodes, a code-agent exact-effect benchmark, manifest-bound ledger faults, proposal-pressure controls, and hosted/API compatibility traces.
  arXiv  Detail & Related papers  (2026-07-07T08:38:50Z)
• Cordon: Semantic Transactions for Tool-Using LLM Agents [11.567293065381918]
  This paper introduces Cordon, a transactional runtime system for staging and validating agent effects before commit.<n>It reduces irreversible-effect failures while preserving benign task completion with modest approval and latency overhead.
  arXiv  Detail & Related papers  (2026-06-16T06:21:14Z)
• LogAct: Enabling Agentic Reliability via Shared Logs [1.3220949653718854]
  We propose a new abstraction called LogAct, where each agent is a deconstructed state machine playing a shared log.<n>LogAct enables agentic introspection, allowing the agent to analyze its own execution history using LLM inference.<n>In our evaluation, LogAct agents recover efficiently and correctly from failures; debug their own performance; optimize token usage in swarms.
  arXiv  Detail & Related papers  (2026-04-09T08:58:23Z)
• AgentSentry: Mitigating Indirect Prompt Injection in LLM Agents via Temporal Causal Diagnostics and Context Purification [25.817251923574286]
  We propose a novel inference-time detection and mitigation framework for large language model (LLM) agents.<n>AgentSentry is the first inference-time defense to model multi-turn IPI as a temporal causal takeover.<n>We evaluate AgentSentry on the textscAgentDojo benchmark across four task suites, three IPI attack families, and multiple black-box LLMs.
  arXiv  Detail & Related papers  (2026-02-26T07:59:10Z)
• CausalArmor: Efficient Indirect Prompt Injection Guardrails via Causal Attribution [49.689452243966315]
  AI agents equipped with tool-calling capabilities are susceptible to Indirect Prompt Injection (IPI) attacks.<n>We propose CausalArmor, a selective defense framework that computes lightweight, leave-one-out attributions at privileged decision points.<n> Experiments on AgentDojo and DoomArena demonstrate that CausalArmor matches the security of aggressive defenses.
  arXiv  Detail & Related papers  (2026-02-08T11:34:08Z)
• ToolSafe: Enhancing Tool Invocation Safety of LLM-based agents via Proactive Step-level Guardrail and Feedback [53.2744585868162]
  Monitoring step-level tool invocation behaviors in real time is critical for agent deployment.<n>We first construct TS-Bench, a novel benchmark for step-level tool invocation safety detection in LLM agents.<n>We then develop a guardrail model, TS-Guard, using multi-task reinforcement learning.
  arXiv  Detail & Related papers  (2026-01-15T07:54:32Z)
• Interact-RAG: Reason and Interact with the Corpus, Beyond Black-Box Retrieval [49.85856484781787]
  We introduce Interact-RAG, a new paradigm that elevates the LLM agent into an active manipulator of the retrieval process.<n>We develop a reasoning-enhanced workflow, which enables both zero-shot execution and the synthesis of interaction trajectories.<n>Experiments across six benchmarks demonstrate that Interact-RAG significantly outperforms other advanced methods.
  arXiv  Detail & Related papers  (2025-10-31T15:48:43Z)
• SAND: Boosting LLM Agents with Self-Taught Action Deliberation [54.48979740613828]
  Large Language Model (LLM) agents are commonly tuned with supervised finetuning on ReAct-style expert trajectories or preference optimization over pairwise rollouts.<n>We propose Self-taught ActioN Deliberation (SAND) framework, enabling LLM agents to explicitly deliberate over candidate actions before committing to one.<n>SAND achieves an average 20% improvement over initial supervised finetuning and also outperforms state-of-the-art agent tuning approaches.
  arXiv  Detail & Related papers  (2025-07-10T05:38:15Z)
• Memory Injection Attacks on LLM Agents via Query-Only Interaction [49.14715983268449]
  We propose a novel Memory INJection Attack, MINJA, without assuming that the attacker can directly modify the memory bank of the agent.<n>The attacker injects malicious records into the memory bank by only interacting with the agent via queries and output observations.<n> MINJA enables any user to influence agent memory, highlighting the risk.
  arXiv  Detail & Related papers  (2025-03-05T17:53:24Z)

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.