Metraipolis › archive › dc_d59834d199a2

Towards a Risk Assessment of Malicious Skill Files in Coding Agents — arXiv:2608.05223

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.05223
License (as archived)
CC0-1.0
Archived
2026-10-27T08:04:31.942Z
SHA-256 of the archived text
a908f09432b57159bd06a2d0cbefc95665a9250da51781c800ffdfc2d30acf0f
Stable URL of this page
https://metraipolis.com/documento/dc_d59834d199a2

How to cite

Towards a Risk Assessment of Malicious Skill Files in Coding Agents — arXiv:2608.05223 — arXiv (en), https://arxiv.org/abs/2608.05223 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d59834d199a2 (Metraipolis archive dc_d59834d199a2, sha256 a908f09432b57159bd06a2d0cbefc95665a9250da51781c800ffdfc2d30acf0f)
BibTeX
@misc{dcd59834d199a2,
  title = {Towards a Risk Assessment of Malicious Skill Files in Coding Agents — arXiv:2608.05223},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.05223},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d59834d199a2; sha256 a908f09432b57159bd06a2d0cbefc95665a9250da51781c800ffdfc2d30acf0f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.05223. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Autonomous coding agents are increasingly embedded in enterprise software workflows with delegated authority over connected systems. Central to this architecture is the agent skills interface: folders of instructions and scripts that agents load dynamically to specialize their behavior. This interface also widens the attack surface, letting malicious shell commands hide within natural-language skill files. We make three contributions. First, an adversarial skill-synthesis method using six LLMs across four families to transform 471 real-world shell commands into benign-appearing skills, released as a benchmark of 2,826 skills mapped to 11 MITRE ATT&CK tactics. Second, a reproducible evaluation pipeline coupling run stratification, evidence anchoring, a refusal veto, and a deterministic declared-intent override with a three-judge LLM-as-a-judge panel, validated against a blind human gold standard (Cohen's kappa = 0.85). Third, a large-scale characterization of two enterprise-grade agents across 5,629 completed runs. Gemini CLI is exploited in 95.5-96.1% of runs and Qwen Code in 71.6-74.0% (raw majority vote to declared-intent-corrected estimate, both within the human gold standard), nearly invariant to the generating model. Explicit safety recognition occurs in only 1.99% of runs. Enterprises must assess and mitigate skill-interface risk before adopting coding agents. Our code and dataset are available at this https URL

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-08-07 in cs.SE).
Categorie arXiv: Software Engineering (cs.SE) ; Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2608.05223 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-27T07:55:47.029Z dalla pagina https://arxiv.org/catchup/cs.SE/2026-08-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).