Reproduce “Evaluating Authorization and Trust Boundaries in Agentic Orchestration Frameworks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sahaya Jestus Lazer, Kshitiz Aryal, Maanak Gupta, Elisa Bertino
The authors evaluate how authorization and information provenance hold across tool-using agents and handoffs in LangChain, LangGraph, AutoGen, and OpenClaw. Using the same model with framework-specific configurations, they test access, delegation, execution, and memory. In unchecked inter-agent handoffs, all four configurations disclosed a protected secret; authorization checks prevented access and disclosure in the tested setups, not as framework guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.