README.md

# Prompt Injection Security Checker

Research-informed, multi-skill guidance for assessing prompt-injection risks in prompts, AI application workflows, retrieval, multimodal inputs, and tool use. This is a skills-only plugin; it includes no scanner, network tool, or guarantee of protection.

Start by defining the system boundary and assets, then trace direct and indirect untrusted input into model context, outputs, and actions. Separate observations from hypotheses, and prioritize mitigations that reduce impact through isolation, least privilege, constrained tools, output/action validation, and human review where warranted. Treat pattern checks and model-based detectors as imperfect layers, not proof of safety.

This plugin provides defensive review and safe test planning. It does not execute attacks against live systems, access private data, exfiltrate information, or perform destructive actions. Any live assessment requires the owner's authorization and a clearly bounded scope; default to static review and harmless test fixtures.

Customer support: https://thedoersfirm.com/support  
Privacy: https://thedoersfirm.com/privacy  
Terms: https://thedoersfirm.com/terms