SEAL: Sealed Queries for Local Sanitization against Indirect Prompt Injection

Summary

SEAL, a method for sanitizing tool outputs against indirect prompt injection in local LLM agents, learns a soft prompt and scoring readout offline and uses a frozen task model. On AgentDojo it reduces attack success from 34.6% to 0.1% and raises joint task completion from 51.4% to 77.5%, using about 127 times fewer defense-related tokens than PISanitizer. The authors also report results on AgentDyn and a second backbone.

Topics

• Prompt injection and jailbreaks

Related items

• InfoUnderstanding and Mitigating Role Spoofing in LLMsSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• HighCVE-2026-119050: Agno PandasTools unsafe reflection allows code execution via prompt injectionSimilar attack · NVD/CVE Database
• Low3 Ways AI Agents Already Operate in Your BrowserSimilar attack · BleepingComputer
• InfoAdversa AI Launches Coding Agent Security Platform to Block Cyberattacks on AI Agents in Real TimeSimilar attack · Adversa AI Blog
• MediumNew Prompt Injection Techniques Cover Agent-Planning Attacks Similar attack · CrowdStrike Blog