In-the-wild manipulation against AI agents in github issues, MCP servers, and skill files.
Aggregate stats from our public scanner across real malicious and benign samples collected from each origin.