Metraipolis › archive › dc_0ef188911224

HarnessSafe: Evaluating Safety Across Persistent Carriers in Agent Harnesses — arXiv:2608.06984

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.06984
License (as archived)
CC0-1.0
Archived
2026-10-19T09:52:59.888Z
SHA-256 of the archived text
11ce8955d3e408c47d590fef83704f5e363da79cbef6e3df1d6d09f3a46bf53c
Stable URL of this page
https://metraipolis.com/documento/dc_0ef188911224

How to cite

HarnessSafe: Evaluating Safety Across Persistent Carriers in Agent Harnesses — arXiv:2608.06984 — arXiv (en), https://arxiv.org/abs/2608.06984 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0ef188911224 (Metraipolis archive dc_0ef188911224, sha256 11ce8955d3e408c47d590fef83704f5e363da79cbef6e3df1d6d09f3a46bf53c)
BibTeX
@misc{dc0ef188911224,
  title = {HarnessSafe: Evaluating Safety Across Persistent Carriers in Agent Harnesses — arXiv:2608.06984},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.06984},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0ef188911224; sha256 11ce8955d3e408c47d590fef83704f5e363da79cbef6e3df1d6d09f3a46bf53c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.06984. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Modern agent harnesses persist state across tasks and sessions through persistent carriers like memory, skills, tools, and shared artifacts. However, this capability creates delayed safety risks: attacker-influenced content can cross system boundaries and later affect the execution of a benign request. Existing benchmarks typically focus on a few carriers or harnesses, while end-to-end attack-success rates reveal little about how risks propagate. To this end, we present HarnessSafe, a benchmark comprising 328 executable cases across seven persistent-carrier families and evaluated on most mainstream agent harnesses. Each case is specified as a Persistent-Risk Lifecycle that traces attacker influence from its initial entry, through persistence across carriers and system boundaries, to a later benign trigger and an observable violation. We further introduce a multi-stage, trace-based evaluation that uses observable execution evidence to determine how far each attack chain progresses and where it is stopped. Experiments show that containment is carrier-specific and strongly depends on the harness-model configuration. Both the harness and model backend substantially shape containment outcomes, while attack success rates cannot reflect distinct lifecycle progression patterns.

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-08-10 in cs.CR).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.06984 — 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 2026-10-19T09:51:05.647Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-08-10?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).