Summary

This repository provides an advanced, autonomous bug bounty agent framework integrating with multiple AI coding assistants like Claude Code, Codex, and Copilot. It automates the security testing lifecycle, from discovery to reporting, enhancing efficiency for researchers through AI-driven hunting and exploit chain building.

How It Works

The framework features ~50 AI agents, 26 commands, and 19 CLI tools, managed by an MCP server. It supports autonomous hunt loops, A→B exploit chain construction, and persistent memory via a "brain" for endpoint tracking and learning. A key component is the optional writeup search, allowing agents to query user-provided FAISS/SQLite indexes for prior art, with a fallback to local payloads. Dynamically translates agents for AI coding tool compatibility.

Quick Start & Requirements

MCP servers launch via uv run --with mcp. Typical setup: clone, set env vars (e.g., HACKERONE_TOKEN), run uv run python3 tools/scaffold.py hackerone tesla, then agent commands. Alternatively, use provider-specific bundles or the tools.installer for global/project installs. Prerequisites: Python 3.10+, uv. Optional: faiss-cpu, sentence-transformers for semantic search. Recommended security tools: nmap, nuclei, etc.

Highlighted Details

• Multi-AI Platform Support: Integrates Claude Code, Codex, Gemini, Cursor, Windsurf, Copilot, and OpenClaw.
• Autonomous Hunting & Exploit Chains: Automated hunting loops and /chain command for A→B→C exploit sequences.
• Writeup Search MCP: Queries user-built FAISS/SQLite indexes for research, falling back to local resources.
• Cost Tracking: Automatic AI token cost logging via CC hooks and live status line display.
• Rigorous Validation: Mandatory 7-Question Gate for findings and platform-specific CVSS version enforcement.
• Scope Enforcement: Pre-ToolUse hook blocks out-of-scope commands.

Maintenance & Community

The README lacks details on community channels, contributors, or sponsorships. The project policy excludes GitHub Actions CI, relying on local pytest.

Licensing & Compatibility

Licensed "For authorized security testing only. Follow responsible disclosure." This restricts usage to legitimate security assessments, potentially limiting commercial use or integration into closed-source projects.

Limitations & Caveats

The advanced writeup search requires users to provide their own FAISS/SQLite index. The restrictive licensing may hinder broader adoption. The absence of automated CI pipelines could be a concern for users prioritizing continuous integration guarantees.