STARS: Skill-Triggered Audit for Request-Conditioned Invocation Safety in Agent Systems
For developers of autonomous language-model agents, this work provides a practical invocation-time risk triage layer that improves detection of unsafe skill invocations, but the improvement is incremental and context-dependent.
The authors propose STARS, a method for real-time risk scoring of skill invocations in agent systems, combining static capability priors with request-conditioned risk models. On a benchmark of 3,000 records, their calibrated fusion achieves 0.439 high-risk AUPRC against indirect prompt injection attacks, outperforming static baselines (0.380) and contextual scorers (0.405), though gains are smaller on in-distribution tests.
Autonomous language-model agents increasingly rely on installable skills and tools to complete user tasks. Static skill auditing can expose capability surface before deployment, but it cannot determine whether a particular invocation is unsafe under the current user request and runtime context. We therefore study skill invocation auditing as a continuous-risk estimation problem: given a user request, candidate skill, and runtime context, predict a score that supports ranking and triage before a hard intervention is applied. We introduce STARS, which combines a static capability prior, a request-conditioned invocation risk model, and a calibrated risk-fusion policy. To evaluate this setting, we construct SIA-Bench, a benchmark of 3,000 invocation records with group-safe splits, lineage metadata, runtime context, canonical action labels, and derived continuous-risk targets. On a held-out split of indirect prompt injection attacks, calibrated fusion reaches 0.439 high-risk AUPRC, improving over 0.405 for the contextual scorer and 0.380 for the strongest static baseline, while the contextual scorer remains better calibrated with 0.289 expected calibration error. On the locked in-distribution test split, gains are smaller and static priors remain useful. The resulting claim is therefore narrower: request-conditioned auditing is most valuable as an invocation-time risk-scoring and triage layer rather than as a replacement for static screening. Code is available at https://github.com/123zgj123/STARS.