InfoSecBulletin Cybersecurity for mankind
PentestAgent, an open-source AI framework by Masic (GH05TCREW), now offers improved features like prebuilt attack playbooks and easy HexStrike integration.
A researcher known as GH05TCREW released a tool on GitHub that uses large language models (LLMs) like Claude Sonnet or GPT-5 through LiteLLM for advanced black-box security evaluations.
Hackers deploy Casbaneiro Trojan that activates on bank websites
German police read Signal, Telegram, WhatsApp messages without breaking encryption
Urgent Patch! cPanel, GitLab Flaws Expose Users to RCE, File and Credential Theft
Palo Alto PAN-OS Flaw Enables Root Arbitrary Code Execution
Critical Check Point VPN flaws allow remote code execution attacks
Cisco confirms CVE-2026-20079 flaw in Secure FMC is exploited in attacks
Hackers exploit PaperCut flaws using hundreds of AI agents, compromising 440 servers globally
CISA Says Chinese Firms Extracted Billions of Tokens From Frontier AI Models
Nightmare Eclipse Drops New Microsoft Defender ‘ShieldCrash’ zero-day
cPanel Flaw Lets Hosting Accounts With Mail Privileges Execute Code as Root
PentestAgent uses a terminal interface with modes for chat assistance, autonomous agents, and multi-agent teams, helping pentesters enhance their work while maintaining control. It’s important to use it legally: only test approved systems, as unauthorized access is illegal.
Features and Playbooks:
PentestAgent comes with its structured attack playbooks, predefined workflows for web app testing like THP3-style assessments. Users launch them via CLI: pentestagent run -t example.com –playbook thp3_web.
These playbooks help the AI with reconnaissance, vulnerability scanning, and exploitation by using domain-specific knowledge from a Retrieval-Augmented Generation (RAG) system.
Session notes categorized as credentials, vulnerabilities, findings, or artifacts are stored in loot/notes.json and create a “shadow graph” in Crew mode, allowing an orchestrator to deploy specialized workers for strategic insights.
The tool supports three operational modes, summarized below:
| Mode | Command | Description |
|---|---|---|
| Assist | (default) | Interactive chat; user directs flow. |
| Agent | /agent <task> | Autonomous single-task execution. |
| Crew | /crew <task> | Multi-agent orchestration. |
