Raptor
Turns Claude Code into a general-purpose offensive/defensive security agent using CLAUDE.md rules, sub-agents, and skill orchestration for adversarial thinking and security operations.
What it does
Open-source project by Gadi Evron that transforms Claude Code into a general-purpose AI security agent. Uses CLAUDE.md configuration files to define rules, sub-agents, and skills that orchestrate security tool usage for adversarial thinking. Enables both offensive operations (attack simulation, vulnerability research) and defensive operations (threat detection, incident analysis). The approach is agent-native: rather than building a security tool that uses AI, it configures an AI agent to become a security practitioner.
Security relevance
Demonstrates how agentic AI can be weaponised for security operations — both offensively and defensively. The sub-agent architecture allows specialised agents (recon, exploit, analysis) to coordinate on complex security tasks. Relevant for understanding how adversaries might use AI agents for attacks, and for building defensive agent-based security workflows.
When to use it
Use for AI-powered red teaming exercises, security research, or building defensive automation. Requires Claude Code access and comfort with agentic AI workflows. Early-stage project — experimental, not production-hardened. Best suited for security researchers exploring the intersection of AI agents and offensive/defensive security.
OWASP coverage
Risks addressed — mapped to both OWASP Top 10 standards. 3 in LLM, 3 in Agentic.
The raw record
What Yuntona stores. Single source of truth — fork it on GitHub.
name: Raptor slug: raptor type: Mixed category: AI Red Teaming url: https://github.com/gadievron/raptor reviewed: 2026-04 added: 2026-04 updated: 2026-04 risks: llm: [LLM01, LLM06, LLM07] asi: [ASI01, ASI04, ASI07] complexity: Plug & Play pricing: — audience: Red Team lifecycle: [test] tags: [Agentic, Claude Code, Offensive, Open Source, Red Teaming]