Prompt Security Gateway (PSG)

🚧 A middleware firewall for LLM/GenAI apps.

Features

• 🔑 Detects jailbreaks, secrets, PII, finance leaks, IP terms
• ⚖️ Severity-based actions: block, redact, log
• 🛡️ Works on inputs & outputs (middleware + post-process)
• 🔌 Extendable via plugins
• 🧑‍🤝‍🧑 Multi-tenant & role-based access
• 📊 Helper functions for rules

Install

npm install prompt-security-gateway

const { createPSG } = require("prompt-security-gateway");

const psg = createPSG({
  onBlock: (reason, payload) => console.warn("🚨", reason, payload),
  rules: [
    { id: "health_0", regex: /\bpatient\s+id:\s*\d+/i, category: "healthcare", severity: "high" }
  ]
});

// Express middleware
app.post("/chat", psg.middleware(), async (req, res) => {
  const { prompt } = req.body;
  // call your LLM here
  const result = await llm(prompt);
  res.json({ result: await psg.postProcess(result) });
});