MoonshotAI
Featured
Kimi-K3
Kimi K3 is an open-weight, 2.8T-parameter native multimodal agentic model with a 1M-token context window, designed for frontier coding, knowledge work, and reasoning tasks.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides a structured, up-to-date taxonomy of the fast-growing field of AI-driven penetration testing, traces the co-evolution of benchmarks and agents, is actively maintained with community contributions, and includes a verified list of commercial systems.
Who it is for
- Cybersecurity researchers and practitioners
- AI and LLM agent developers
- Penetration testers and red-team operators
- Graduate students in cybersecurity and AI
Use cases
- Conducting literature reviews on automated penetration testing
- Identifying research gaps and trends in LLM-driven offensive security
- Benchmarking new systems against the provided taxonomy and corpus
- Evaluating commercial pentesting platforms against academic work
Strengths
- Comprehensive corpus of 81 papers systematically categorized into 6 research categories
- Clearly defined four-phase architectural evolution from text-only reasoning to reinforcement learning
- Actively maintained with pull requests and additions beyond the original survey
- Includes a verified list of 8 commercial systems for industry context
Considerations
- Survey corpus is fixed to papers published between 2023 and June 2026, potentially missing later work
- Not all listed papers have publicly available code or datasets
- Taxonomy assigns each paper to a single primary category, which may oversimplify multi-faceted contributions
README quick start
A Survey of LLM-Driven Penetration Testing: Taxonomy, Co-Evolution, and Open Challenges
Agent4Pentest: A Curated Paper List and Survey Companion
English | 简体中文
Survey corpus: 81 papers · 6 research categories · 2023–June 2026
Agent4Pentest is a curated paper list for LLM-driven and agent-based penetration testing and the official companion repository for A Survey of LLM-Driven Penetration Testing: Taxonomy, Co-Evolution, and Open Challenges.
We continuously maintain this collection, add newly released work, and welcome researchers to contribute their papers, benchmarks, systems, datasets, and code through Pull Requests.
[!NOTE] The survey analyzes a fixed corpus of 81 works released between 2023 and June 2026. This repository is a living index and may grow beyond the original survey corpus as the field evolves.
Table of Contents
- Research Landscape
- Paper Collection
- Evaluation Benchmarks
- General-Purpose Systems
- Domain-Specific Frameworks
- CTF-Based Systems
- Defense-Oriented Research
- Surveys
- Commercial Agents & Platforms
- Contributing
- Citation
Research Landscape
The survey maps Agent4Pentest through a six-category taxonomy and a four-phase architectural evolution, then relates this progression to the parallel expansion of benchmark and CTF-based training infrastructure.
Four-Phase Architectural Evolution
| Phase | Core shift | Main bottleneck |
|---|---|---|
| I. Text-only Reasoning (2023) | LLMs reason over the engagement state while humans execute every command. | Execution autonomy and high human dependence |
| II. Tool-augmented Single Agents (2023–2024) | A single agent directly invokes scanners, exploit frameworks, and shells. | Context management and reasoning degradation on long tasks |
| III. Multi-agent Coordination (2024–2025) | Specialized subagents split the attack pipeline under an orchestrator, enabling structured handoffs and parallel execution. | Training-data scarcity and dependence on human demonstrati |
Related repositories
Similar projects matched by category, topics, and programming language.
xuchonglang
Featured
investing-for-beginners
A structured investing guide for Chinese beginners covering US stocks, options, and cryptocurrency, with focus on foundational concepts and risk awareness.
Krishnagangwal
Featured
CS-Fundamentals
A curated collection of Computer Science fundamentals (PDFs, notes, cheatsheets, interview question banks) for placement preparation, covering seven core subjects plus general resources.