In progressSince Sep 2026University of Edinburgh

Safety of LLM Agent Harnesses

Overview

Evaluating the safety of language models deployed inside agent harnesses.

This work is in progress. Details and results will be shared here once it is published. Get in touch if you'd like to discuss it.

Results

Results will be posted here once the work is published.

Updates

Sep 2026Project started.

Cite

Work in progress. If you refer to this project before a paper is available, please cite this page:

@misc{saxena2026harnesssafety,
  title        = {Safety of {LLM} Agent Harnesses},
  author       = {Saxena, Rohit},
  year         = {2026},
  howpublished = {\url{https://saxenarohit.github.io/projects/harness-safety/}},
  note         = {Work in progress}
}