It is argued that this path from a candidate action to privileged side effects must be governed outside the model, and study an architecture interposing four roles (planner, policy gate, executor, auditor) between agent and operating system.
Abstract
Large language model agents are increasingly granted real privileges (executing commands, modifying files, calling APIs), so an agent that errs has already acted. Existing defences concentrate on the agent's inputs, while the path from a candidate action to privileged side effects remains less directly studied. We argue that this path must be governed outside the model, and study an architecture interposing four roles (planner, policy gate, executor, auditor) between agent and operating system. Two choices are central: actions arrive as structured intents, so adjudication never parses shell syntax; and approval is a one-shot credential bound to the exact bytes that will run. We evaluate on a 313-case benchmark across eight variants, with prompt-only baselines from three hosted LLMs on 150 stratified cases, five repetitions (2,250 attempted calls; 2,249 completed). Effective attack success falls from 98.3% under direct execution to 7.7% deployed. Re-execution against the real implementation yields a similar aggregate rate (7.6% over 66 sandbox-evaluable payloads) but substantial case-level disagreement, at a corrected false-denial rate of 11.1%. A substantial final-stage reduction (from 30.8% to 7.7%) is attributable to the operating-system sandbox, and the benchmark found four implementation defects, none by design review.
Agent skills are Markdown documents that a tool-using agent reads as instructions and then acts on with the privileges of the session that loaded them. Installing one grants authority (the standing to direct the agent, not new privilege), and the size of that grant is what we measure. We present a two-phase attack-surf...
Ratnesh Pandey, Japneet Singh, Harsh Akshit et al.· 0 citations
A tool-using model can follow a malicious instruction even when its credentials are valid. We study whether task-scoped authorization contains the resulting tool execution. Our paired-replay testbed samples a model request once and submits the same action, resource, and arguments to broad bearer, scoped JWT, sender-con...
Language-model agents commonly collapse two decisions that security systems separate: selecting an action and determining whether that action is authorized. When a model exercises standing authority, indirect prompt injection can convert influence over its reasoning into unauthorized effects. We present AgentCap, a run...
Enterprise AI agents can execute altered versions of approved actions, concurrently reuse the same permission, or retry operations whose responses were lost. These failures arise when approval and resource execution lack a shared contract. KedgeFlow specifies language-independent contracts for trusted controllers to au...
Kairos is presented, a reference monitor at the isolation boundary of an unmodified, opaque agent, where a tool call, a shell command, and a script the agent writes all cross the boundary as the same file and network operations, so that confinement holds regardless of the agent's behavior.
Kristoffer Hempel, Nils Loose, Lorenzo Cavallaro et al.· 0 citations
It is argued that this gap leaves an irreducible residual of undetected attacks wherever inputs and actions are not restricted in advance to an enumerated set, and systematize defenses across runtime monitoring, architectural separation, and authorization.
Li Zhang, Yang Sun, Jie Shi· arXiv.org· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.