What Makes a Good LLM Agent for Real-world Penetration Testing?
About
LLM-based agents show promise for automating penetration testing, yet reported performance varies widely across systems and benchmarks. We analyze 28 LLM-based penetration testing systems and evaluate five representative implementations across three benchmarks of increasing complexity. Our analysis reveals two distinct failure modes: Type A failures stem from capability gaps (missing tools, inadequate prompts) that engineering readily addresses, while Type B failures persist regardless of tooling due to planning and state management limitations. We show that Type B failures share a root cause that is largely invariant to the underlying LLM: agents lack real-time task difficulty estimation. As a result, agents misallocate effort, over-commit to low-value branches, and exhaust context before completing attack chains. Based on this insight, we present Excalibur, a penetration testing agent that couples strong tooling with difficulty-aware planning. A Tool and Skill Layer eliminates Type A failures through typed interfaces and retrieval-augmented knowledge. A Task Difficulty Assessment (TDA) mechanism addresses Type B failures by estimating tractability through four measurable dimensions (horizon estimation, evidence confidence, context load, and historical success) and uses these estimates to guide exploration-exploitation decisions within an Evidence-Guided Attack Tree Search (EGATS) framework. Excalibur achieves up to 91% task completion on CTF benchmarks with frontier models (39 to 49% relative improvement over baselines) and compromises 4 of 5 hosts on the GOAD Active Directory environment versus 2 by prior systems. These results show that difficulty-aware planning yields consistent end-to-end gains across models and addresses a limitation that model scaling alone does not eliminate.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Web security task completion | XBOW 104 tasks | Task Completion Rate91 | 32 | |
| Active Directory domain escalation | GOAD 5 hosts | Domain Escalation Success Rate80 | 30 | |
| End-to-end penetration testing | PentestGPT Ben 13 machines | Machines Rooted12 | 30 | |
| Penetration Testing | HTB Easy Season 8 | Completed Attempts4 | 1 | |
| Penetration Testing | HTB Medium Season 8 | Completed Count4 | 1 | |
| Penetration Testing | HTB Hard Season 8 | Completed Count2 | 1 | |
| Penetration Testing | HTB Insane Season 8 | Completions0.00e+0 | 1 | |
| Penetration Testing | HTB Season 8 (Total) | Completed10 | 1 |