Reproduce “SkillMarker: A KGW-Based Watermarking Method for Agent Skill Ownership Verification” 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

Alan Tang, Lifeng Wang, Gang Wu, Tao Zhang, Yantao Jia, Xiaofeng Zhu

SkillMarker adapts KGW watermarking to agent skills by protecting operational details and rewriting flexible wording to carry a keyed signal. It uses constrained candidate generation and search, then detects the signal in suspect text. Across four datasets, task pass-rate changes ranged from −2.01% to +0.93%; tested contiguous 1,536-token copies were detected in specified backgrounds, not universally.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.