New Open Source Framework Automates Bug Bounty Testing with 50 Specialized Agents

The open source Pentest Agent Suite brings a validator gate and persistent memory tracker to automate vulnerability discovery across seven AI coding platforms.

CSBadmin
3 Min Read

Framework Architecture

A new open source tool called Pentest Agent Suite has been released on GitHub, designed to automate bug bounty hunting through a collection of 50 specialized security agents. Developed by researcher H-mmer, the framework operates across seven major AI coding platforms including Claude Code, OpenAI Codex, Google Gemini, Cursor, Windsurf, VS Code Copilot, and OpenClaw. It includes 26 slash commands, 19 CLI tools, and a cross-IDE installer that generates native configuration files for each supported environment.

The suite is built around three core layers: the agent collection, a dual-server Model Context Protocol infrastructure, and a comprehensive rules library. The bounty platforms MCP server integrates with 16 bug bounty programs such as HackerOne, Bugcrowd, Intigriti, Immunefi, and YesWeHack. It exposes seven tools for platform interaction, including listing platforms, retrieving program scope, syncing programs, drafting reports, and submitting findings. A separate writeup search MCP server uses FAISS semantic search, SQLite keyword search, or a zero dependency local fallback to query a bundled rule file containing over 2,600 lines covering attack patterns like XSS, SSRF, SQLi, IDOR, OAuth, SSTI, JWT, and others.

Impact and Automated Validation

The framework introduces a validation pipeline called the 7 Question Gate, run by a dedicated validator agent on every finding. If any question receives a negative answer, the finding is automatically killed, downgraded, or flagged for additional chaining. No finding can be submitted without passing a validation check and achieving a quality score of at least 7 out of 10. The autopilot mode enforces multi layer encoding in every payload attempt and refuses to declare an attack surface exhausted until a full exhaustion matrix is complete, with configurable paranoid, normal, or yolo checkpoint modes.

A persistent tracking component monitors every endpoint per target and enforces circuit breaker logic: five consecutive 403 or 429 responses trigger a 60 second auto backoff. It syncs cross engagement knowledge through incremental hash based diffing, preventing redundant testing across sessions. The installer generates native configuration formats for each supported tool and writes them to the appropriate IDE directories, making deployment straightforward across different development environments.

Source: Cyber Security News

CSBadmin

The latest in cybersecurity news and updates.

Share This Article
Follow:
The latest in cybersecurity news and updates.