Security for the agent layer, not the model layer.
AI red-team tooling: prompt-injection PoCs, adversarial testing, honeypots, passive threat detection, agent guardrails, and trust infrastructure for the MCP ecosystem.
| Repo | What it does |
|---|---|
| Red-Agent | AI-powered proxy command center — natural-language red-team operations |
| AI-PASSWORD-TESTING-AGENT | AI-driven password strength testing and vulnerability research |
| Ghost-Reaper | AI-powered full-spectrum passive threat detection |
| MIRAGE | AI cyber range & honeypot simulator for red-team practice |
| SCAFFOLD-WATCH | Silent guardian for Claude Code — real-time builder monitoring |
| STEGANO | Invisible-Unicode injection attacks against LLMs + detection PoC |
| mcp-trust-registry | Agent-pointable MCP registry with trust verification |
- 𝕏 @Insider747 — launch threads, PoC demos, red-team notes