Metraipolis › archive › dc_19b91fcd0c3f

When Tool Outputs Become Commands: Separating Action Induction from Runtime Authorization in Tool-Augmented LLM Agents — arXiv:2608.27146

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.27146
License (as archived)
CC0-1.0
Archived
2026-10-17T14:23:31.897Z
SHA-256 of the archived text
be77ed1320c819e11fa46ee5aa153347bd159044cb3e528556641ccd7d343332
Stable URL of this page
https://metraipolis.com/documento/dc_19b91fcd0c3f

How to cite

When Tool Outputs Become Commands: Separating Action Induction from Runtime Authorization in Tool-Augmented LLM Agents — arXiv:2608.27146 — arXiv (en), https://arxiv.org/abs/2608.27146 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_19b91fcd0c3f (Metraipolis archive dc_19b91fcd0c3f, sha256 be77ed1320c819e11fa46ee5aa153347bd159044cb3e528556641ccd7d343332)
BibTeX
@misc{dc19b91fcd0c3f,
  title = {When Tool Outputs Become Commands: Separating Action Induction from Runtime Authorization in Tool-Augmented LLM Agents — arXiv:2608.27146},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.27146},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_19b91fcd0c3f; sha256 be77ed1320c819e11fa46ee5aa153347bd159044cb3e528556641ccd7d343332},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27146. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Tool-augmented LLM agents must rely on untrusted runtime Observations to complete open-ended tasks; however, when tool outputs no longer merely provide data but begin to specify concrete actions, they effectively become ``commands'' that can drive real-world side effects beyond user intent. We argue that this risk arises from conflating action induction with execution authorization. To address this distinction, we propose SARA, which treats action induction and execution authorization as distinct runtime roles and separates action provenance from execution authority. On the Observation side, a context-isolated Action Probe exposes action-inducing semantics and persistently records action-origin provenance across steps as a review signal; on the execution side, actual tool calls are authorized only against the user objective and audited evidence from authorized successful executions, while satisfying goal, execution-chain, and argument-level support. To preserve this separation across multi-step execution, SARA applies No-History-Promotion to prevent historical recurrence from laundering action origins into execution authority. Across AgentDojo and AgentDyn, SARA limits ASR to no more than \(0.63\%\) across four primary evaluation settings while maintaining competitive task utility, and consistently reduces ASR across additional Agent backbones.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-28 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Software Engineering (cs.SE).
Pagina del lavoro: https://arxiv.org/abs/2608.27146 — 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 2026-10-17T13:16:53.552Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).