Preprint
Aug 2026
How Do LLM Agents Actually Get the Flag? Trace-Level Provenance for Agentic Offensive Security Evaluation
CTF-ABACUS is introduced, a trace-based agent auditing framework that reconstructs each run as an evidence-grounded solve profile that provides a basis for designing benchmarks that better isolate the offensive capabilities of autonomous language-model agents.
Kimberly Milner, Minghao Shao, Nanda Rani et al.
· 0 citations