How Agentic AI Coding Assistants Become the Attacker's Shell

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Liu, Yue, Zhao, Yanjie, Lyu, Yunbo, Zhang, Ting, Wang, Haoyu, Lo, David
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910255310438400
author Liu, Yue
Zhao, Yanjie
Lyu, Yunbo
Zhang, Ting
Wang, Haoyu
Lo, David
author_facet Liu, Yue
Zhao, Yanjie
Lyu, Yunbo
Zhang, Ting
Wang, Haoyu
Lo, David
contents Agentic AI coding assistants can edit files, run commands, and access the internet on behalf of developers. However, their reliance on unvetted external artifacts introduces a new attack vector. Hidden instructions in external artifacts can hijack these assistants, turning them into an attacker's shell to run unauthorized commands. In this article, we examine how these prompt injection attacks work, measure their prevalence, discuss the limitations and challenges of current defenses, and suggest future research directions.
format Preprint
id arxiv_https___arxiv_org_abs_2605_25871
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle How Agentic AI Coding Assistants Become the Attacker's Shell
Liu, Yue
Zhao, Yanjie
Lyu, Yunbo
Zhang, Ting
Wang, Haoyu
Lo, David
Software Engineering
Cryptography and Security
Agentic AI coding assistants can edit files, run commands, and access the internet on behalf of developers. However, their reliance on unvetted external artifacts introduces a new attack vector. Hidden instructions in external artifacts can hijack these assistants, turning them into an attacker's shell to run unauthorized commands. In this article, we examine how these prompt injection attacks work, measure their prevalence, discuss the limitations and challenges of current defenses, and suggest future research directions.
title How Agentic AI Coding Assistants Become the Attacker's Shell
topic Software Engineering
Cryptography and Security
url https://arxiv.org/abs/2605.25871