🔐 What Can Go Wrong with Agents
1️⃣ Perception - Prompt injection, poisoned data, fake documents, malicious payloads
2️⃣ Reasoning - Hallucinations, bias, unsafe logic, policy violations
3️⃣ Planning - Unsafe actions, privilege escalation, illegal sequences
4️⃣ Action - Data leaks, fraud, destructive actions
5️⃣ Reflection = Poisoned memory, corrupted learning, policy drift
2️⃣ Reasoning - Hallucinations, bias, unsafe logic, policy violations
3️⃣ Planning - Unsafe actions, privilege escalation, illegal sequences
4️⃣ Action - Data leaks, fraud, destructive actions
5️⃣ Reflection = Poisoned memory, corrupted learning, policy drift
