🛡️ Spec: Cloudflare Security Audit (AI-AND-LLM) & OWASP LLM Aligned
Verify AI Agent Skills & MCP Tools
Before They Touch Your Infrastructure
Audit SKILL.md, system prompts and MCP servers against indirect prompt injection, tool-sink code execution, credential exfiltration, and unbounded delegated loops.
AgentShield will pull SKILL.md or the specified file directly and run full heuristic static analysis against Cloudflare AI-AND-LLM spec.