Stop Common Prompt Injection Phrases (v1) in Hugging Face Traffic
AI security governance · MCP security · LLM gateway
What it does
Keyword-based detection of frequently observed prompt-injection phrases. v2 (ML-backed) is planned in the future-tasks file.
How Igris secures Hugging Face traffic
Igris Content Guard inspects Hugging Face requests and responses for Common Prompt Injection Phrases (v1) using keywords-based detection, then blocks matches.
Gateway config for Hugging Face
// Igris gateway → Hugging Face
baseUrl: "https://api-inference.huggingface.co/v1"
contentGuard:
detectors: ["openai-injection-v1"]
action: "deny"Other detectors for Hugging Face
Secure your AI estate with Igris