ATR (Agent Threat Rules) — request to include in AI Security Solutions Landscape

[eeee2345]

Requesting inclusion of ATR in the AI Security Solutions Landscape for Agentic AI.

Tool: ATR (Agent Threat Rules)
Type: Open-source detection rules standard (regex-based, like Snort/Sigma but for AI agents)
License: MIT
Website: https://agentthreatrule.org
Repository: https://github.com/Agent-Threat-Rule/agent-threat-rules

Coverage:

• 108 detection rules across 9 threat categories
• OWASP Agentic Top 10: 10/10 coverage
• SAFE-MCP: 78/85 techniques (91.8%)
• Benchmark: 96.9% recall (SKILL.md), 99.7% precision (PINT MCP)
• 53,577 skills scanned, 0% FP on clean content

Adoption:

• Cisco AI Defense ships 34 ATR rules in production (feat: Add ATR community rule pack — 34 rules for MCP tool poisoning, multi-agent attacks, and advanced injection cisco-ai-defense/skill-scanner#79)
• NVIDIA Garak integration under review (detectors: add Agent Threat Rules NVIDIA/garak#1676)
• npm: 23K+ downloads/month

Lifecycle stage: Detection / Runtime Monitoring
Paper: https://doi.org/10.5281/zenodo.19178002

We also have an open PR (#814) on the LLM Top 10 project adding ATR detection examples to all ASI01-ASI10 categories.

Happy to provide any additional information needed for the landscape evaluation.