🛡️ AI Application Threat Modeling & Prompt Injection Audit

A comprehensive security audit, architecture threat model, and adversarial red-teaming assessment of an LLM-powered enterprise application.

---

📌 Audit Highlights

• Frameworks Used: OWASP Top 10 for LLM Applications, STRIDE Threat Modeling
• Attack Vectors Tested: Direct Prompt Injection, System Prompt Leakage, Indirect Payload Injection
• Target Architecture: RAG-enabled Customer Support Agent with Function Calling APIs

---

📂 Repository Contents

• THREAT_MODEL.md: Full architectural risk mapping and mitigation matrix.
• RED_TEAM_RESULTS.md: Step-by-step logs of prompt injection tests and exploitation proofs.

---

🛠️ Recommended Security Architecture

To secure LLM pipelines against prompt injection and excessive agency, implement a Dual-LLM Guardrail Architecture:]