I build security tools for AI agents, working at the intersection of Generative AI and Cybersecurity. Focus: making autonomous AI agents safer — from prompt injection detection to EU AI Act compliance scanning.
Key numbers: 225 detection patterns · 15 languages · F1=98.3% · <6ms latency · OWASP LLM+Agentic+MCP Top 10: 100% coverage · 33 security advisories filed (285k+ Stars reach)
🛡️ ClawGuardSecurity Scanner for AI Agents
Real-time detection of prompt injection, jailbreaks, code injection, social engineering and data exfiltration. 225 patterns across 12 categories in 15 languages. Zero LLM dependency, <6ms latency. |
Security Scanning REST API · Live →
REST API for AI security: 225 patterns, PDF compliance reports mapped to EU AI Act (Art. 9, 15, 17, 61). Free tier, Pro & Enterprise plans. Stripe payments, Docker-deployed. |
|
CI/CD gate: scan prompts and MCP configs on every PR. Configurable severity thresholds, PR annotations, fail-on-critical. |
MCP server for Claude Desktop, Cursor & Windsurf — security scanning directly in your editor. 5 tools, FastMCP + async. |
| Repo | Description |
|---|---|
| 🐍 Shield Python SDK | pip install clawguard-shield — 3 lines to scan any prompt. Dataclass results, type hints, FastAPI & LangChain ready. |
| 📦 Shield JS SDK | npm install clawguard-shield — Zero deps, ESM+CJS+TS, full TypeScript support. |
| 🧪 Prompt Lab | Interactive Prompt Injection Playground — test 225 patterns in real time. Live → |
| 📊 Project Manager | Project dashboard with Gantt, Kanban & early warning system — Next.js 16, React 19, Tailwind CSS 4 |
- 33 security advisories filed across major open source projects (Google genai-toolbox, Notion MCP, Laravel, n8n, GhidraMCP, git-mcp, XcodeBuildMCP and more)
- 285k+ combined Stars across affected repositories
- Blog: 14 articles on prompt injection, MCP security, EU AI Act compliance
- Registry Dashboard: 11,529 MCP servers scanned for security vulnerabilities
Building the security layer between AI agents and the real world.